> We funded the open-source SQLite VFS shim that helped isolate the race condition almost immediately, and will help track down similar bugs in the future.
Interesting example of a company funding open source - in this case paying for the development of a new and very specific debugging tool.
Not only are they funding open source, they are actively allowing users to host their own control plane via headscale which is a libre implementation of the tailscale control protocol and developed by an engineer who works for Tailscale by day. This instantly made me trust and like them, even if at first I was cautious because I naturally mistrust anything that gets a lot of hype. I've been running headscale on NixOS, hosting my own tailnet since last year and everything just works out of the box. Tailscale lives up to the hype! <3
> Because this wouldn’t be a quick or easy fix, we reached out to the SQLite developers for a professional support contract. This was a great decision. It gave us direct access to their deep expertise and experience, and we had many detailed technical conversations about our architecture and our incidents.
They were willing to pay to get help solving the problem, and then pay again to make sure that the problem is easier to avoid in the future! That kind of long-term thinking seems pretty rare nowadays...
I wish they'd buy the carcass of Keybase from Zoom.
It seems very similar to Tailscale: immense utility from a free-tier product for the general public, which leads to trust and a large enterprise market.
In real life, an imaginary "leadership" quickly evaporates upon the first hurdle of any kind until proven otherwise by real actions. If I was in your shoes, I would be very careful with blanket statements like these.
Reading the article, it sounds more like they funded this by buying a SQLite support contract, and the SQLite developers created this in the process of helping them track down the issue.
> A single Go process exclusively accesses that database, and serves the control plane for those tailnets. This single-writer design is exactly how SQLite is meant to be used.
This line led me to believe that the writer and checkpointing logic lived on the same database connection, so I was curious to find out how the data race occurred. However, the bug details on the SQLite page[0] outline that it can only ever occur if there are multiple connections open, so the writer and the checkpointer must have been on different threads.
Database corruption is due to 2+ peer connects with 744 file permission entering header rwxr -.- WAL write new content into secondary header tag: inter-element whitespace.
> Multiple processes can have the same database open at the same time. Multiple processes can be doing a SELECT at the same time. But only one process can be making changes to the database at any moment in time, however.
I don't know enough about the scale of Tailscale's operations to comment strongly on this, but if they're fairly significant shouldn't that have read "is exactly how MariaDB is meant to be used" or "exactly how Postgres is meant to be used"? SQLite has a "lite" in the name for a reason, but it's often pushed into places where it's being asked to do things it was never really designed for.
I would not think of SQLite as "lite" anything. It's SQL In The Executable.
It has a better security and data-durability track record than both Postgres and MySQL, and often beats them in the sorts of things applications do with databases:
This particular bug doesn’t seem to arise from SQLite’s “lite” nature. It’s a TOCTOU inside the DB when applying WAL segments in a checkpoint, which is a pattern used in extremely similar ways by Postgres and MySQL. They don’t seem to have similar bugs, but I don’t think there’s any reason to believe that this is due to their being client/server rather than coordinated-file databases.
It says a lot about sqlite that a bug becomes front-page news on HN. I'm impressed that Tailscale took this seriously enough to engage with a commercial support contract. I'd love to work for a company that cared so much about correctness.
I admit to curiosity as to whether static analysis could have caught this. E.g., Rust's type system (yeah yeah I know) catches all data races, unless they originate in unsafe code, which this one might or might not have; a hypothetical Rust SQLite would probably need a lot of unsafe (https://github.com/tursodatabase/turso has 556 unsafe blocks in the core), and I don't have a sense of whether the particular part that contained this bug would be included in that.
It sounds like it was a bug caused by pushing on performance/safety ratio, it could easily have been avoided by adding enough mutexes in the code, but for performance reasons (I assume) they don't do that. To write it like that in Rust you'd have to use unsafe blocks and then you don't gain much from the typesystem in this case.
Very nice article, and I appreciate SQLite's explanation of the bug too. And how extremely cool Tailscale appears to have been about it (paying for the VFS shim, etc.).
I'd have liked to have heard more about the decision to checkpoint so frequently that put them on this path though. Presumably that's to keep the WAL tiny for very fast recovery. Trying to mitigate some of the deleterious effects of inserting a DBMS into your network layer, I suppose? Tricky stuff. Wonder how that compares to typical etcd snapshot frequencies too.
Great read. So glad they took the time to tell this story. (And glad they, as a for profit corporation, took out a support contract with SQLite. I hope they continue to do so even though this problem is resolved.)
This was really, really interesting - what a triumphant adventure.
A few (very, very, very pedantic) things that stood out:
> We wanted a way to restore service that didn’t involve rolling back to the last known-good backup (which would lose a lot of data) or repairing the known-corrupted database (which was potentially risky).
(Emphasis mine) - it would be "risky", not "potentially risky" - then the "calculated risk period" starts and it's "potentially problematic".
In the SQLite report[0] (11.2) I wish they downplayed this less - a mention of the rarity, then technical details - I'm friendly with a few of the devs/previous-devs, have the utmost respect for their skill and accomplishments (and by extension, faith that the developers I do not personally interact with are also excellent), appreciation and fondness for the huge accomplishment that is SQLite, and on and on... this is world-class work. Maybe section 11.2 wasn't really aimed at me, or I'm too critical. To be fair to all involved, what a minor quibble for such an interesting problem/fix. I hope my comment isn't a fly in the ointment.
Last bugfix point[1] - ugh. What a sinking feeling that must've been to deploy a fix then be flooded with not-green - and a lesson[2] against smuggling other changes in a changeset "just because we're already here"? Happy it turned out non-catastrophic, but did result in a rare (not remembering other instances of top of head) recall[3] from SQLite. That it was throwing errors at the same time SQLite and Tailscale were testing the other WAL-issue bug must've upset some stomachs for a moment.
>In our control plane, we take manual control of the checkpoint process so we can run fast and consistent backups.
> running boring technology in a non-standard way is a risk.
It was a good read and reminder that the industry is loosing experts gradually. I am not a DBA and yet I have heard about this behavior at least couple times in the past as something to avoid. Its just one of those things which didnt get a chance to be documented cause experts avoided it and regulars didn't get into
I've never heard of any reliability reason you shouldn't run checkpoints whenever you want - only performance reasons. Can you elaborate?
Backing up sqlite by copying the file (e.g. rsync) while it's open is a surefire way to eventually get corruption caused by a race condition, but it seems like tailscale wasn't doing that. They were probably using the proper sqlite backup API.
But you don't need checkpoints for consistency and I think the backup API will not copy both the old and new versions of pages just because they're in the WAL, in other words I think checkpointing then backing up should give you the same pages as backing up without checkpointing. So the whole thing seems unnecessary.
My generalist reading of this (which is applicable in lot of cases) is that a dedicated flow is handling the house keeping part of the job, aka resetting the checkpoint after writing the WAL to disk. These one off house keepers are common in a lot of softwares and they expect to work alone, they are tested to work alone. They always have a set of ritual, rules and order in which they do all the things. Now if some other thread takes off some of those jobs then they break this routine for the house keeping job and this new thread/person may not always know what else has to be done before and after this one particular job for the sake of completeness.
On second question of why the tailscale developers did it, its possibly for the same reason why they invested this much into debugging this issue. Some one believed the current behavior did not fit into their architecture, they want to be more performant and take control over things. A big part of me considers this is a required exercise to try, grow and learn. The only thing they could have for improvement would be to have these old hands on architect kind of folks on their team who might have hinted/pointed them to the problem a long before. Challenge/chances are that these older folks would have even stopped them from going in this direction in the design phase itself.
If you already know the bug is caused by a race between writing and checkpointing then it's easy. I'm unsure how the linked article makes your point at all though.
Block device upfuckery layers are powerful against databases. Years ago some colleagues wrote one that provides most of the hazards described by "Parity Lost and Parity Regained"[1] to test FoundationDB, which immediately uncovered several flaws in a project that described itself as well-tested. It's easy to do this with all the probing features that Linux (and others) provide today.
With opus 4.7-ish to fable 5, immediately after release - before they locked it down, it was shockingly easy to find crash bugs in a lot of very heavily used DBs and other software.
See perhaps recent video "Reliability Lessons From SQLite - Richard Hipp | SSW 2026":
> Abstract: SQLite is a C-language library that implements a self-contained, in-process relational database engine supporting full-featured SQL, an advanced query planner, and ACID transactions. By many estimates, SQLite is the most widely used software library in the world today.
> Over its 26-year history, SQLite has gained a reputation as software that "just works". This talk goes over the design choices and development practices that have, at least in the opinion of the lead developer, resulted in that reputation.
Great writeup, and it was great to see them step in an pay the developers of SQLite to help them fix the bug. I get tired of corporations asking open source authors to fix problems that affect the corporation for free. And while I'm sure it was frustrating for folks to have these outages, I find such puzzles pretty fun to get to the bottom of.
As a simple user of SQLite, I think this level of debugging is incredible and appreciate being a beneficiary of the ecosystem and hard work of others. Thank you!
As others have said: great article! I did find myself wanting them to get to the point, but once they started describing the bug and the fix, it was very satisfying. I'm very happy there are companies out there on the frontiers of functionality not only funding fixes and debugging measures, but taking the time to write up the details so we can all benefit.
Tailscale just moved up in my priorities list. Was going to host my next website with hostinger, but now I'm going to at least try to run a personal server with tailscale to make it public. I might not be able to figure it all out, and may end up going with the VPS route, but this gave me some appreciation for the company that makes me willing to try the less familiar method.
The irony is that the SQLite developers get a support contract iff someone runs off the path in anger and finds an ancient bug. But perhaps that's part of what make it a quality team: devotion thriving without adverse incentives.
While you might be correct, I wouldn't necessarily assume the "and only if" part of your statement. There might be companies that choose to proactively purchase support contracts. And there are companies that have paid $150K/year for https://sqlite.org/consortium.html access. Which means, among other things, that they get first priority for any needs they have:
> Consortium members have the guaranteed, undivided attention of the SQLite developers for 23 staff-days per year and for as much additional time above and beyond that amount that the core developers have available. There are no arbitrary limits on contact time. The consortium will never be over-subscribed. New SQLite developers will be recruited and trained as necessary to cover the 23 day/year support commitment.
The SQLite home page lists five companies that have paid for consortium access. I can easily imagine that there are more who don't want to pay $150K/year but would pay $1.5k/year, proactively, to get "private, expert email advice from the developers of SQLite" when they need it.
The fun part is unless your app is pretty high traffic you might see the bug only once, or twice, and never have a reasonable way to even get close to a fix or know what your real exposure is.
That is the correct mentality, but sometimes, you’re wrong. I found (not entirely true; someone else [0] found it first, but my report from my company got traction [1] on it) a weird bug in ProxySQL several months ago involving mirroring and fast routing, where if you had a configuration that was illogical from the standpoint of documented behavior, it would duplicate queries, which led to super fun times for writes.
I spent hours empirically trying various scenarios before concluding that no, it was a ProxySQL bug.
I mean, ProxySQL and SQLite are at very different tiers of reliability. I encountered multiple unreported data-corrupting (and some resource exhausting/connection mis-pinning) bugs in ProxySQL within a few months of using it for the first time, and I wasn’t using it for anything particularly complex or advanced—just a basic connection pool, no failover or caching/rewriting/replica awareness, but a lot of frontends and QPS.
It gives me a warm feeling when companies invest in open source support in this way. Helping great projects get even better is somehow better than releasing yet another project.
Normal code has 50% to 90% ratio of code coverage by unit-tests. Dynamic-typed languages (Python, Ruby) usually require more, like 100% - 120%.
SQLite has 59,000% ratio [1]
Yet it didn't help for a bug to left unnoticed for 16 years :( I don't know what we can do for the industry. I doubt one can formally verify a project like SQLite, and keep it maintainable.
One might be able to formally verify the sqlite pager layer, where this problem occurred. It's just writing pages to disk and reading pages from disk, and providing ACID guarantees.
This reaffirms my belief that SQLite is not well suited for systems with significant concurrency. It replaces fopen, not postgres. Although this corruption is a rare bug and sqlite is usually extremely stable, it's usually not worth it from a performance and features standpoint either.
Here they were trying to do a backup by forcing a checkpoint and then copying the file. Systems like postgres let you do online continuous backups.
Eeeh, this particular bug was a race when applying WAL to the data files in a checkpoint. Postgres’s checkpointing is theoretically just as vulnerable to this class of bugs as SQLite, though it doesn’t seem to have equivalent issues today.
Online backups/replicas are nice until you’re charged for network traffic or have to recreate the replica from scratch, at which point the initial-restore-then-hook-up-the-WAL-stream dance is prone to all sorts of racy issues. If you’re lucky enough to have only a single process talking to the DB, SQLite seems like a nice way to sidestep that complexity while keeping a simple backup story.
Heck, this is basically the Redis model: a single process/thread coordinates all access to the data, and occasionally forks off a background job to snapshot the state somewhere. From that perspective, the Tailscale controller binary is a database; SQLite is just the data file format.
Yeah, I also feel 'just use SQLite [no matter what]' is just the pendulum swinging hard after the 'just use mongodb [no matter what]' of yesteryear.
It's so sometimes just performative. I remember when Tailscale had a similar performative approach with 'just use a JSON file on disk'. Then etcd. Then SQLite. Like sure, you can keep picking the absolite mininum technology for your needs and then change it every couple of years... Or you could just immediately go with a solid Postgres (or Postgres-like setup, eg. yugabyte) and save yourself a bunch of faffing about with weird solutions and migrating between them. But I guess that doesn't drive engagement on your blog.
Oh wow. I checked your site, but I (still) don't see the ability for average joes (and their pet AI agents) to signup. Could be huge, even if just offered in 5 minute slots. Is your email contact still valid in your profile? Want to reach out (Moose is the name). FYI, got a job at Antithesis a while ago after some rigorous interviewing.
The sentence you’ve quoted actually links to a list of ways to corrupt SQLite, which I think is interesting in its own right: https://www.sqlite.org/howtocorrupt.html
I believe your flaky SD card is category 4, Disk Drive and Flash Memory Failures.
I have encountered this exactly once, and it was in fact a flaky SD card - running a small web service off of a raspberry pi, with the SQLite DB stored on the SD.
Checked and this 100% compatible SQLite3 database (also C-API) did not contain the bug:
https://github.com/punnerud/mpedb
(Disclaimer: My own project)
And this statement is wrong in the article:
“ Because SQLite is a single-writer database with serialisable transactions, our transaction history was completely linear and deterministic. (This wouldn’t be true in a multi-writer database like Postgres or MySQL.)”
Actually possible in mpedb to replay multi-writer, and actually better than SQLite3. Try to reply now() in a statement, that is not deterministic in SQLite but is in MPEdb.
Maybe it's just me, but the explanations of the cause don't align.
One clue was that during corruption incidents, our metrics showed that SQLite would report copying more pages from the WAL file than were actually available. If there are 10 pages in the WAL file and 20 pages get copied to the database, something is clearly wrong.
vs
it thinks some of the pages have been copied from the WAL into the main database file, but they haven’t. Those pages never get written to the database file, and that data is permanently lost.
The first says "more were copied than existed" but the second says "fewer were copied than should have been."
Like I said, it's probably just me interpreting something incorrectly.
I haven't looked into the actual code fix, but given the "reset" name I have to think it has to do with SQLite "thinking" it has copied more pages than it actually did.
i.e. The checkpoint starts, and a write hits after the modifications to data structures have been done but before the data has actually been put in the database. The process starts over again, but doesn't undo the changes it made to indexes etc. Hence the db thinks it holds pages that don't exist.
Those seem consistent to me. Some pages weren't written to the WAL (yet?), but something else referenced them or otherwise indicated they existed, so then the other process tried to read them, resulting in the issue in the first quote.
Same bug, two angles. Nothing really copies 20 pages, the "20" is just a broken counter. SQLite's internal tally of how many WAL pages it already saved to the main db gets corrupted and reads too high. Since it trusts that tally, it assumes those pages are already saved and skips writing them for real, so when the WAL resets they're gone. The "copied more than existed" number is the bug showing in your metrics; "pages never written" is the actual damage.
Glad this got found and fixed, but I continue to be astounded at the amount of work people put into making SQLite do things that would be much simpler with other systems.
Maybe, but there’s nothing about this particular bug that’s due to SQLite not being a networked database. Postgres is just as likely to have TOCTOU races in its checkpointer, which works roughly the same way.
Tailscale likely deals with a lot of security-sensitive traffic, given the nature of the service. I'm guessing one of the requirements they were given was to have a tiny blast radius in case encryption keys got leaked, and that meant isolating each customer's tailnet (meta)data to it's own sqlite db rather than letting everyone share a postgres cluster.
That assumes that they do actually believe it's a mistake. They didn't explain the reasons they've gone for this architecture in much/any detail. I'd be interested in hearing them talk more about that in the future.
Perhaps you or I would make a different decision based on the aims that lead them there. But there isn't enough information to say whether or not their decision was a mistake, even if it has lead to a peculiar bug. It may be perfectly legitimate and we just don't know some of the constraints they had.
Richard Hipp's recent talk at Software Should Work explains that AI agents have been testing SQLite and they've gotten a deluge of new bug reports from the fuzz-like testing they can do. But they do not do this in house; hobbyists and other organizations do this in their own internal agent-driven fuzzing.
huh. ran into almost-certainly this, but blamed it on litestream and rearchitected a bit as a result. will have to see if I can reproduce the issue as we were using with the patched sqlite
I see tailscale also use the pure Go SQLite conversion so I hope this fix will land there soon, it's rapidly become one of my favorite packages for self-contained tools.
Such a good write up . Having explored a little bit of sqlite internals for a codecrafters challenge i was mildly happy i could follow along what was happening .
I would assume they actually moved off of litestream here no? otherwise how can their frequent manual checkpointing even succeed when litestream locks for the same behavior.
> This investigation is a useful reminder: running boring technology in a non-standard way is a risk. The common paths and standard configurations are incredibly well-tested and reliable. Most people use SQLite in a standard configuration and never face this sort of issue. Everything we were doing was a public, documented, supported configuration—but by taking manual control of the checkpointing process and running at our own aggressive pace, we stepped off the well-trodden operational path.
I feel like they missed a key takeaway from their own argument here, they should not be running a non-standard configuration :)
We have very good reasons for our checkpointing model, related to our backup + disaster recover strategy, along with resource cost. It might be worth writing about one day, so I'll not give away all the details, but in very short form, we organize a backup strategy that has minimal pause time, avoids doubling the page cache cost of the database, and enables extremely fast byte-copy restores in disaster recovery.
> Whenever corruption occurred, we had to stop the control plane process on the shard while we repaired or restored the database. This was painful for tailnets on that shard, because their entire control plane disappeared during that recovery window.
This is maybe one of the purest examples of the Bell Curve Meme in software engineering. The things you would do to the system Tailscale operates to eliminate all single points of failure (generally, and in the specific case where, where the "single point of failure" applies only to a small cohort of customers) would make the system less resilient, and increase failures.
Generally, you do complex distributed systems without on-paper single-points-of-failure anywhere when you absolutely have to, because those systems don't have transient failures. That's not mesh networks like Tailscale at all.
You don’t need the control plane most of the time. I had a zero downtime headscale upgrade because once the nodes negotiate through the control plane they can talk to each other all the time. The data plane is peer to peer.
It’s problematic because you can’t run connections but it doesn’t stop the world.
Well it's SPoF for a shard not the whole thing so in practice that makes this considerably better. Secondly, it's not a datapath SPoF so presumably all tailscaled processes would continue running and exchanging traffic no problem - you just won't be able to launch any new ones.
What are some solutions to avoid database corruption being single points of failure? I can’t think of any off the top of my head. I don’t think people typically consider database corruption to be a kind of failure common enough to design for, unless you have unusual requirements.
The general answer to this is Byzantine consensus, which cryptocurrency blockchains are designed to solve. If your nodes are willing to fail a little more politely (e.g. no lying, immediately crashing, etc) you can use something cheaper like raft/paxos.
But yeah, it's a lot cheaper to build a reliable system than it is to be resilient.
This is a great example of outages looking different from the perspective of the operator vs the user. Because there's many shards the blast radius of failure is contained to a small subset of users but for those users it's an outage. The way it's designed you can't lose any shards without impacting users. Compare to say Elasticsearch where it's possible to lose nodes and lose shards without the user noticing. One approach isn't universally better than the other.
I lost my Android SMS DB way back in the day (20+ years? dang I'm getting old) because the SMS app's fix when opening the DB detected any issue was to delete it and start fresh.
> Nobody wanted us to spend six months looking for bugs in SQLite
Most companies wouldn't. Instead, they'd fire the weirdo that came up with the idea of using some weird db, and switch to Postgres like God intended.
I'm not saying either is right, this is not a criticism of Tailscale and their approach, paying the Sqlite maintainers to fix a real bug is commendable, but it certainly doesn't inspire confidence in the "Sqlite in production" hype train.
While Sqlite is indeed boring technology for single-user SQL DBs, for traditional CRUD and network services, Postgres seems to be a much better trodden and much safer path.
We wanted a way to restore service that didn’t involve rolling back to the last known-good backup (which would lose a lot of data) or repairing the known-corrupted database (which was potentially risky). To do this, we built a
transaction logging pipeline.
We streamed every SQL statement that modified the database to a separate log file. Because SQLite is a single-writer database with serializable transactions, our transaction history was completely linear and
deterministic.
(This wouldn’t be true in a multi-writer database like Postgres or MySQL.) Replaying those transactions against the latest known-good backup should restore the database to its most recent state, safely bypassing the corruption.
[...] This pipeline worked, but then it did something even better: it gave us a clue.
[...] To understand what was happening during these faulty checkpoints, the SQLite developers created a new debugging tool for the virtual filesystem layer.
[...] To help diagnose our problem, the SQLite developers created a wrapper around the virtual filesystem that
writes additional tracing information and logs
about changes to the database.
[...] After our next corruption incident, the additional logs from the new tmstmpvfs shim allowed the SQLite developers to find and fix the bug:
a rare data race
in the SQLite source code between a checkpoint and a write transaction."
Great article!
Software Engineering lessons (that repeat in this article!): So called "Heisenbugs" (bugs that make it past developer test harnesses and a company's Quality Assurance (QA) team) that show up post-deployment intermittently and can't be reproduced locally, occur because one or more of the following factors:
1) The lack of Determinism in a software process or processes.
2) The lack of appropriate logging.
3) The lack of the ability to replay a software process, step by exact step, state by exact state, as it has occurred in the field (occurs as an effect of #1 and/or #2).
4) Multi-threaded code; i.e., multiple threads giving rise to race conditions or other very specific intermittent combinatorial/permutational conditions caused by multiple threads and specific sections of code, which due to very large numbers of permutational timing possibilities, were not or could not be exactly tested for in development...
Anyway, great article! A must-read for any Sr. Software Engineer, or any developer that wrestles with hard-to-find-and-fix bugs in the field...
While technically true as written, it seems to downplay the significance:
> The bug is a data race with tight timing constraints. It is unlikely to occur in common use.
A large customer did experience this corruption, so it's important for people with tailscale's setup update immediately.
> The developers have never been able to reproduce the bug organically and had to add special testing logic to SQLite that deliberately triggers the circumstances of the the bug in order to verify that the issue has been fixed.
Odd not to highlight the sentence where they answer the obvious question "Why Tailscale in particular?":
> They also explained why we were more likely to hit the bug than other SQLite users: we take manual control of the checkpointing process, and we checkpoint very aggressively. Even a bug triggered by a rare condition was bound to hit us eventually.
The bug requires two checkpoints in very quick succession, which presumably isn't something sqlite would do on its own, as it would be a pointless waste of performance.
Quick note that data corruption bugs that are impossible to reproduce are not uncommon (perhaps they're the norm). So some amount of head scratching trying to figure out a plausible scenario by which the system could get into the state represented by the smoking remains is often required. Then you attempt to force it into the supposed bad state by modifying code paths accordingly. So the approach used in this case is clever, but it's not particularly unusual in the world of data stores.
This was a great technical writeup and very interesting to read, but it's not clear to me why once the suspected source of the bug was identified, they seemingly didn't build a automated way to trigger the condition? It seems like that could have cut down on the uncertainty of whether the fix worked over a painfully long period of time.
> It could exist that long because it was rare—so rare, the SQLite developers had to add code to deliberately trigger it in their testing environments.
Yes of course I mean the tailscale team. Having an independent way to repro the bug besides waiting for it to happen in prod seems pretty basic eng best practice.
> We funded the open-source SQLite VFS shim that helped isolate the race condition almost immediately, and will help track down similar bugs in the future.
Interesting example of a company funding open source - in this case paying for the development of a new and very specific debugging tool.
Not only are they funding open source, they are actively allowing users to host their own control plane via headscale which is a libre implementation of the tailscale control protocol and developed by an engineer who works for Tailscale by day. This instantly made me trust and like them, even if at first I was cautious because I naturally mistrust anything that gets a lot of hype. I've been running headscale on NixOS, hosting my own tailnet since last year and everything just works out of the box. Tailscale lives up to the hype! <3
Important features like App Connectors won't work with HeadScale.
Unfortunately, HeadScale is not a drop in replacement for TailScale. Most people don't realize this until they start self hosting.
Due to this, I had to migrate from Tailscale to NetBird, which is completely open source.
https://avilpage.com/2026/06/moving-from-tailscale-to-netbir...
3 replies →
The only nit on my end is that you have to configure a few settings on client and server to avoid sending diagnostics to TS.
For macOS, you need to build the open source tailscale client. Or create file in esoteric file location to opt out using App Store.
For iOS I don’t even think it’s possible to opt out.
1 reply →
Yeah, tailscale seems to have leadership with their head on right, I agree with the way they handle a lot of things.
Yeah, this part also stuck out to me:
> Because this wouldn’t be a quick or easy fix, we reached out to the SQLite developers for a professional support contract. This was a great decision. It gave us direct access to their deep expertise and experience, and we had many detailed technical conversations about our architecture and our incidents.
They were willing to pay to get help solving the problem, and then pay again to make sure that the problem is easier to avoid in the future! That kind of long-term thinking seems pretty rare nowadays...
31 replies →
I wish they'd buy the carcass of Keybase from Zoom.
It seems very similar to Tailscale: immense utility from a free-tier product for the general public, which leads to trust and a large enterprise market.
7 replies →
Their CEO is a very nice and personable guy too. Has given me and others advice on random topics of his interest with no nonsense plenty of times.
Avery is one of the few people I have enough respect for to look up to.
1 reply →
At least for now. All it takes is one greedy executive decision to stop supporting headscale. Given it’s an American company, it’s entirely possible.
1 reply →
In real life, an imaginary "leadership" quickly evaporates upon the first hurdle of any kind until proven otherwise by real actions. If I was in your shoes, I would be very careful with blanket statements like these.
1 reply →
That's actually SQLite revenue model.
https://sqlite.org/prosupport.html
Reading the article, it sounds more like they funded this by buying a SQLite support contract, and the SQLite developers created this in the process of helping them track down the issue.
Started using them like 2 weeks back, happy to see how they work.
New? The sqlite VSF shim is much older. Nice spin.
Spin?
Looks to me like this is the tool in question, added in January: https://github.com/sqlite/sqlite/commits/master/ext/misc/tms...
Well written post, really enjoyed reading it.
> A single Go process exclusively accesses that database, and serves the control plane for those tailnets. This single-writer design is exactly how SQLite is meant to be used.
This line led me to believe that the writer and checkpointing logic lived on the same database connection, so I was curious to find out how the data race occurred. However, the bug details on the SQLite page[0] outline that it can only ever occur if there are multiple connections open, so the writer and the checkpointer must have been on different threads.
[0] https://sqlite.org/wal.html#the_wal_reset_bug
Database corruption is due to 2+ peer connects with 744 file permission entering header rwxr -.- WAL write new content into secondary header tag: inter-element whitespace.
Bug details:
[0]:https://sqlite.org/wal.html#the_wal_reset_bug
> The bug only affects databases in WAL mode when there are two or more database connections open on the same file, in separate threads or processes
To be honest, I'm surprised that someone using SQLite would try to access it directly from multiple threads or processes without fear of data racing.
> Multiple processes can have the same database open at the same time. Multiple processes can be doing a SELECT at the same time. But only one process can be making changes to the database at any moment in time, however.
https://sqlite.org/faq.html#q5
One writer, multiple readers is a specifically supported way of using SQLite.
Why should you be worried if it is used as designed?
2 replies →
[flagged]
I don't know enough about the scale of Tailscale's operations to comment strongly on this, but if they're fairly significant shouldn't that have read "is exactly how MariaDB is meant to be used" or "exactly how Postgres is meant to be used"? SQLite has a "lite" in the name for a reason, but it's often pushed into places where it's being asked to do things it was never really designed for.
> SQLite has a "lite" in the name for a reason
I would not think of SQLite as "lite" anything. It's SQL In The Executable.
It has a better security and data-durability track record than both Postgres and MySQL, and often beats them in the sorts of things applications do with databases:
https://sqlite.org/speed.html
> it's often pushed into places where it's being asked to do things it was never really designed for
https://sqlite.org/whentouse.html
https://sqlite.org/hirely.html
Seems like it absolutely is "designed" for this use case.
> SQLite has a "lite" in the name for a reason
It"s actually SQL "ite" as in rocks, minerals and fossils. Their version control system is called "Fossil".
This particular bug doesn’t seem to arise from SQLite’s “lite” nature. It’s a TOCTOU inside the DB when applying WAL segments in a checkpoint, which is a pattern used in extremely similar ways by Postgres and MySQL. They don’t seem to have similar bugs, but I don’t think there’s any reason to believe that this is due to their being client/server rather than coordinated-file databases.
It says a lot about sqlite that a bug becomes front-page news on HN. I'm impressed that Tailscale took this seriously enough to engage with a commercial support contract. I'd love to work for a company that cared so much about correctness.
Indeed, SQLite has got to be one of the best tested pieces of software with famously 100% test coverage.
https://sqlite.org/testing.html
It's amazing that a bug could exist for 16 years but it is sobering.
I think FoundationDB has that crown. https://antithesis.com/company/about/
SQLite: 92 million lines of tests
Dijkstra: Tests can only prove the presence of bugs, never their absence!
Everyone knows that tests don't prevent all bugs. But they are very good at preventing known bugs from recurring in the future.
Everyone doesn’t seem to know that, because tests are often cited as a way to ensure that AI-generated code is correct.
3 replies →
Donald Knuth: Beware of bugs in the above code; I have only proved it correct, not tried it.
The perfect yang to Dijkstra's yin
I admit to curiosity as to whether static analysis could have caught this. E.g., Rust's type system (yeah yeah I know) catches all data races, unless they originate in unsafe code, which this one might or might not have; a hypothetical Rust SQLite would probably need a lot of unsafe (https://github.com/tursodatabase/turso has 556 unsafe blocks in the core), and I don't have a sense of whether the particular part that contained this bug would be included in that.
There was a post recently using TLA to model this issue and (re)discover the bug: https://news.ycombinator.com/item?id=48730953.
4 replies →
I don't think they could. This is probably a memory-mapped file, which would be unsafe.
It sounds like it was a bug caused by pushing on performance/safety ratio, it could easily have been avoided by adding enough mutexes in the code, but for performance reasons (I assume) they don't do that. To write it like that in Rust you'd have to use unsafe blocks and then you don't gain much from the typesystem in this case.
I wouldn't think so, but Go also has a data race detector: https://go.dev/doc/articles/race_detector.
1 reply →
>Rust's type system (yeah yeah I know)
What do you know?
1 reply →
It can prove absence of specific bugs though.
no it can't. what if this bug is still there but the timing window is now one CPU instruction?
10 replies →
You simply don't have enough testcases.
Testing:
The union of a lot of necessary conditions is not a sufficient condition. But it might be good enough for software.
The idea of testing is to sample and test the paths you care about, mostly business workflows, not to enumerate infinite combinations.
It's an art to come up with a great test suite that covers just enough and minimizes overlap, not only survives but also helps with refactoring.
probatio diabolica
https://en.wikipedia.org/wiki/Probatio_diabolica
https://07th-expansion.fandom.com/wiki/Devil%27s_Proof
https://news.ycombinator.com/item?id=18442941
Jesus Christ. I have this comment saved, but it gets me every time.
Obligatory UNIX fortune: "Computer Science can be thought of as the post-Turing decline in formal methods."
Very nice article, and I appreciate SQLite's explanation of the bug too. And how extremely cool Tailscale appears to have been about it (paying for the VFS shim, etc.).
I'd have liked to have heard more about the decision to checkpoint so frequently that put them on this path though. Presumably that's to keep the WAL tiny for very fast recovery. Trying to mitigate some of the deleterious effects of inserting a DBMS into your network layer, I suppose? Tricky stuff. Wonder how that compares to typical etcd snapshot frequencies too.
Agreed, I was also wondering about this. Maybe the aggressive checkpointing was for preventing WAL-overflow?
Was wondering the same, doesn’t mention if they tried checkpointing less frequently
Great read. So glad they took the time to tell this story. (And glad they, as a for profit corporation, took out a support contract with SQLite. I hope they continue to do so even though this problem is resolved.)
This was really, really interesting - what a triumphant adventure.
A few (very, very, very pedantic) things that stood out:
> We wanted a way to restore service that didn’t involve rolling back to the last known-good backup (which would lose a lot of data) or repairing the known-corrupted database (which was potentially risky).
(Emphasis mine) - it would be "risky", not "potentially risky" - then the "calculated risk period" starts and it's "potentially problematic".
In the SQLite report[0] (11.2) I wish they downplayed this less - a mention of the rarity, then technical details - I'm friendly with a few of the devs/previous-devs, have the utmost respect for their skill and accomplishments (and by extension, faith that the developers I do not personally interact with are also excellent), appreciation and fondness for the huge accomplishment that is SQLite, and on and on... this is world-class work. Maybe section 11.2 wasn't really aimed at me, or I'm too critical. To be fair to all involved, what a minor quibble for such an interesting problem/fix. I hope my comment isn't a fly in the ointment.
Last bugfix point[1] - ugh. What a sinking feeling that must've been to deploy a fix then be flooded with not-green - and a lesson[2] against smuggling other changes in a changeset "just because we're already here"? Happy it turned out non-catastrophic, but did result in a rare (not remembering other instances of top of head) recall[3] from SQLite. That it was throwing errors at the same time SQLite and Tailscale were testing the other WAL-issue bug must've upset some stomachs for a moment.
[0] https://sqlite.org/wal.html#the_wal_reset_bug
[1] https://tailscale.com/blog/sqlite-wal-reset-bug#fixed-with-a...
[2] Nobody conceptually learned anything here - we're all just reminded of what we know: that sometimes "perfect storms" do actually occur.
[3] https://sqlite.org/releaselog/3_52_0.html
>In our control plane, we take manual control of the checkpoint process so we can run fast and consistent backups.
> running boring technology in a non-standard way is a risk.
It was a good read and reminder that the industry is loosing experts gradually. I am not a DBA and yet I have heard about this behavior at least couple times in the past as something to avoid. Its just one of those things which didnt get a chance to be documented cause experts avoided it and regulars didn't get into
I've never heard of any reliability reason you shouldn't run checkpoints whenever you want - only performance reasons. Can you elaborate?
Backing up sqlite by copying the file (e.g. rsync) while it's open is a surefire way to eventually get corruption caused by a race condition, but it seems like tailscale wasn't doing that. They were probably using the proper sqlite backup API.
But you don't need checkpoints for consistency and I think the backup API will not copy both the old and new versions of pages just because they're in the WAL, in other words I think checkpointing then backing up should give you the same pages as backing up without checkpointing. So the whole thing seems unnecessary.
My generalist reading of this (which is applicable in lot of cases) is that a dedicated flow is handling the house keeping part of the job, aka resetting the checkpoint after writing the WAL to disk. These one off house keepers are common in a lot of softwares and they expect to work alone, they are tested to work alone. They always have a set of ritual, rules and order in which they do all the things. Now if some other thread takes off some of those jobs then they break this routine for the house keeping job and this new thread/person may not always know what else has to be done before and after this one particular job for the sake of completeness.
On second question of why the tailscale developers did it, its possibly for the same reason why they invested this much into debugging this issue. Some one believed the current behavior did not fit into their architecture, they want to be more performant and take control over things. A big part of me considers this is a required exercise to try, grow and learn. The only thing they could have for improvement would be to have these old hands on architect kind of folks on their team who might have hinted/pointed them to the problem a long before. Challenge/chances are that these older folks would have even stopped them from going in this direction in the design phase itself.
In other words there exists a concept of HOT and COLD backups for this reason only.
For this particular category of bug, SQLite's existing testing methodology is demonstrably outclassed by modern deterministic concurrency testing.
https://antithesis.com/blog/2026/wal-reset-bug/
If you already know the bug is caused by a race between writing and checkpointing then it's easy. I'm unsure how the linked article makes your point at all though.
Block device upfuckery layers are powerful against databases. Years ago some colleagues wrote one that provides most of the hazards described by "Parity Lost and Parity Regained"[1] to test FoundationDB, which immediately uncovered several flaws in a project that described itself as well-tested. It's easy to do this with all the probing features that Linux (and others) provide today.
1: https://www.usenix.org/legacy/event/fast08/tech/full_papers/...
With opus 4.7-ish to fable 5, immediately after release - before they locked it down, it was shockingly easy to find crash bugs in a lot of very heavily used DBs and other software.
I'll accept crashing in preference to a database acknowledging transaction that weren't durably committed.
1 reply →
Did you personally find them?
I'm definitely going to use the word "upfuckery" instead of fault injection the next time I need it.
See perhaps recent video "Reliability Lessons From SQLite - Richard Hipp | SSW 2026":
> Abstract: SQLite is a C-language library that implements a self-contained, in-process relational database engine supporting full-featured SQL, an advanced query planner, and ACID transactions. By many estimates, SQLite is the most widely used software library in the world today.
> Over its 26-year history, SQLite has gained a reputation as software that "just works". This talk goes over the design choices and development practices that have, at least in the opinion of the lead developer, resulted in that reputation.
* https://www.youtube.com/watch?v=V_qzqY1bb7I
Great writeup, and it was great to see them step in an pay the developers of SQLite to help them fix the bug. I get tired of corporations asking open source authors to fix problems that affect the corporation for free. And while I'm sure it was frustrating for folks to have these outages, I find such puzzles pretty fun to get to the bottom of.
As a simple user of SQLite, I think this level of debugging is incredible and appreciate being a beneficiary of the ecosystem and hard work of others. Thank you!
As others have said: great article! I did find myself wanting them to get to the point, but once they started describing the bug and the fix, it was very satisfying. I'm very happy there are companies out there on the frontiers of functionality not only funding fixes and debugging measures, but taking the time to write up the details so we can all benefit.
Tailscale just moved up in my priorities list. Was going to host my next website with hostinger, but now I'm going to at least try to run a personal server with tailscale to make it public. I might not be able to figure it all out, and may end up going with the VPS route, but this gave me some appreciation for the company that makes me willing to try the less familiar method.
I don't think Tailscale can do that though? It can't open a port from the internet to your private network.
It can, but with some specifics about DNS names and supported ports: https://tailscale.com/docs/features/tailscale-funnel
https://tailscale.com/docs/reference/tailscale-cli/funnel
The irony is that the SQLite developers get a support contract iff someone runs off the path in anger and finds an ancient bug. But perhaps that's part of what make it a quality team: devotion thriving without adverse incentives.
While you might be correct, I wouldn't necessarily assume the "and only if" part of your statement. There might be companies that choose to proactively purchase support contracts. And there are companies that have paid $150K/year for https://sqlite.org/consortium.html access. Which means, among other things, that they get first priority for any needs they have:
> Consortium members have the guaranteed, undivided attention of the SQLite developers for 23 staff-days per year and for as much additional time above and beyond that amount that the core developers have available. There are no arbitrary limits on contact time. The consortium will never be over-subscribed. New SQLite developers will be recruited and trained as necessary to cover the 23 day/year support commitment.
The SQLite home page lists five companies that have paid for consortium access. I can easily imagine that there are more who don't want to pay $150K/year but would pay $1.5k/year, proactively, to get "private, expert email advice from the developers of SQLite" when they need it.
What a brutal bug. I'd never entertain a that bug in SQLite could be causing problems in code I wrote.
The fun part is unless your app is pretty high traffic you might see the bug only once, or twice, and never have a reasonable way to even get close to a fix or know what your real exposure is.
That is the correct mentality, but sometimes, you’re wrong. I found (not entirely true; someone else [0] found it first, but my report from my company got traction [1] on it) a weird bug in ProxySQL several months ago involving mirroring and fast routing, where if you had a configuration that was illogical from the standpoint of documented behavior, it would duplicate queries, which led to super fun times for writes.
I spent hours empirically trying various scenarios before concluding that no, it was a ProxySQL bug.
[0]: https://github.com/sysown/proxysql/issues/2233
[1]: https://github.com/sysown/proxysql/pull/5385
I mean, ProxySQL and SQLite are at very different tiers of reliability. I encountered multiple unreported data-corrupting (and some resource exhausting/connection mis-pinning) bugs in ProxySQL within a few months of using it for the first time, and I wasn’t using it for anything particularly complex or advanced—just a basic connection pool, no failover or caching/rewriting/replica awareness, but a lot of frontends and QPS.
It gives me a warm feeling when companies invest in open source support in this way. Helping great projects get even better is somehow better than releasing yet another project.
Which SQLite driver for Go does Tailscale use?
https://github.com/tailscale/sqlite
Normal code has 50% to 90% ratio of code coverage by unit-tests. Dynamic-typed languages (Python, Ruby) usually require more, like 100% - 120%.
SQLite has 59,000% ratio [1]
Yet it didn't help for a bug to left unnoticed for 16 years :( I don't know what we can do for the industry. I doubt one can formally verify a project like SQLite, and keep it maintainable.
https://sqlite.org/testing.html
One might be able to formally verify the sqlite pager layer, where this problem occurred. It's just writing pages to disk and reading pages from disk, and providing ACID guarantees.
This reaffirms my belief that SQLite is not well suited for systems with significant concurrency. It replaces fopen, not postgres. Although this corruption is a rare bug and sqlite is usually extremely stable, it's usually not worth it from a performance and features standpoint either.
Here they were trying to do a backup by forcing a checkpoint and then copying the file. Systems like postgres let you do online continuous backups.
Eeeh, this particular bug was a race when applying WAL to the data files in a checkpoint. Postgres’s checkpointing is theoretically just as vulnerable to this class of bugs as SQLite, though it doesn’t seem to have equivalent issues today.
Online backups/replicas are nice until you’re charged for network traffic or have to recreate the replica from scratch, at which point the initial-restore-then-hook-up-the-WAL-stream dance is prone to all sorts of racy issues. If you’re lucky enough to have only a single process talking to the DB, SQLite seems like a nice way to sidestep that complexity while keeping a simple backup story.
Heck, this is basically the Redis model: a single process/thread coordinates all access to the data, and occasionally forks off a background job to snapshot the state somewhere. From that perspective, the Tailscale controller binary is a database; SQLite is just the data file format.
Yeah, I also feel 'just use SQLite [no matter what]' is just the pendulum swinging hard after the 'just use mongodb [no matter what]' of yesteryear.
It's so sometimes just performative. I remember when Tailscale had a similar performative approach with 'just use a JSON file on disk'. Then etcd. Then SQLite. Like sure, you can keep picking the absolite mininum technology for your needs and then change it every couple of years... Or you could just immediately go with a solid Postgres (or Postgres-like setup, eg. yugabyte) and save yourself a bunch of faffing about with weird solutions and migrating between them. But I guess that doesn't drive engagement on your blog.
Postgres has also had pretty serious bugs, eg fsyncgate
1 reply →
Was curious so we checked and yep, Antithesis finds this bug in about 15 minutes. Will post a repro/writeup here soon.
RECEIPTS: https://antithesis.com/blog/2026/wal-reset-bug/
Oh wow. I checked your site, but I (still) don't see the ability for average joes (and their pet AI agents) to signup. Could be huge, even if just offered in 5 minute slots. Is your email contact still valid in your profile? Want to reach out (Moose is the name). FYI, got a job at Antithesis a while ago after some rigorous interviewing.
We are working on getting self-serve signups to you ASAP. For now, just email us. We don’t bite.
Discussion here: https://news.ycombinator.com/item?id=49277799
Wait, it's AI? AI probably has information about this bug by now, since it was discovered and fixed in March.
1 reply →
I'm equal parts intrigued and skeptical- I guess if the prompt doesn't lead on there is a bug there then I'm impressed.
Stay tuned!
> SQLite corruption is possible, but it’s highly unusual and not something you should encounter in normal operation
If there's a hardware failure, for example a flaky SD card, it's not out of the question. A mobile app with a lot of usage will see it.
(Yes, I know this appears to be a server use case.)
The sentence you’ve quoted actually links to a list of ways to corrupt SQLite, which I think is interesting in its own right: https://www.sqlite.org/howtocorrupt.html
I believe your flaky SD card is category 4, Disk Drive and Flash Memory Failures.
I have encountered this exactly once, and it was in fact a flaky SD card - running a small web service off of a raspberry pi, with the SQLite DB stored on the SD.
Checked and this 100% compatible SQLite3 database (also C-API) did not contain the bug: https://github.com/punnerud/mpedb (Disclaimer: My own project)
And this statement is wrong in the article: “ Because SQLite is a single-writer database with serialisable transactions, our transaction history was completely linear and deterministic. (This wouldn’t be true in a multi-writer database like Postgres or MySQL.)”
Actually possible in mpedb to replay multi-writer, and actually better than SQLite3. Try to reply now() in a statement, that is not deterministic in SQLite but is in MPEdb.
At least mention that it's your own project ...
Added, thanks
Maybe it's just me, but the explanations of the cause don't align.
One clue was that during corruption incidents, our metrics showed that SQLite would report copying more pages from the WAL file than were actually available. If there are 10 pages in the WAL file and 20 pages get copied to the database, something is clearly wrong.
vs
it thinks some of the pages have been copied from the WAL into the main database file, but they haven’t. Those pages never get written to the database file, and that data is permanently lost.
The first says "more were copied than existed" but the second says "fewer were copied than should have been."
Like I said, it's probably just me interpreting something incorrectly.
I haven't looked into the actual code fix, but given the "reset" name I have to think it has to do with SQLite "thinking" it has copied more pages than it actually did.
i.e. The checkpoint starts, and a write hits after the modifications to data structures have been done but before the data has actually been put in the database. The process starts over again, but doesn't undo the changes it made to indexes etc. Hence the db thinks it holds pages that don't exist.
That's my interpretation, anyways.
Those seem consistent to me. Some pages weren't written to the WAL (yet?), but something else referenced them or otherwise indicated they existed, so then the other process tried to read them, resulting in the issue in the first quote.
The explanation says they were written to the WAL and then not written to the main database file.
My interpretation is that they haven't been copied because they didn't exist?
If you have 10 pages and it tries to copy 20, either those 10 pages wouldn't really be copied, or bogus data would be written.
That's how I read at least. Those things are not mutually exclusive.
Or you could have 10 pages, it actually copies 9, and reports 20 anyways.
1 reply →
Same bug, two angles. Nothing really copies 20 pages, the "20" is just a broken counter. SQLite's internal tally of how many WAL pages it already saved to the main db gets corrupted and reads too high. Since it trusts that tally, it assumes those pages are already saved and skips writing them for real, so when the WAL resets they're gone. The "copied more than existed" number is the bug showing in your metrics; "pages never written" is the actual damage.
Glad this got found and fixed, but I continue to be astounded at the amount of work people put into making SQLite do things that would be much simpler with other systems.
Maybe, but there’s nothing about this particular bug that’s due to SQLite not being a networked database. Postgres is just as likely to have TOCTOU races in its checkpointer, which works roughly the same way.
Tailscale likely deals with a lot of security-sensitive traffic, given the nature of the service. I'm guessing one of the requirements they were given was to have a tiny blast radius in case encryption keys got leaked, and that meant isolating each customer's tailnet (meta)data to it's own sqlite db rather than letting everyone share a postgres cluster.
How about having devices generate and not share their private keys, instead?
4 replies →
Yeah it’s a pity that wasn’t addressed in the article.
This is a little like “we shot ourselves in the foot and then performed surgery on our foot, and everything is resolved now.”
That assumes that they do actually believe it's a mistake. They didn't explain the reasons they've gone for this architecture in much/any detail. I'd be interested in hearing them talk more about that in the future.
Perhaps you or I would make a different decision based on the aims that lead them there. But there isn't enough information to say whether or not their decision was a mistake, even if it has lead to a peculiar bug. It may be perfectly legitimate and we just don't know some of the constraints they had.
1 reply →
> Now we’re in summer, we’re confident that we’ve found the bug, that we understand it—and more importantly, that we’ve fixed it.
This is the feeling I chase as a software engineer. It's the greatest motivator.
I imagine the SQLite eschews AI generated code, but using it for testing (vulnerability, performance, etc) would seem like an easy win.
I know their proprietary testing framework is their secret sauce so we may never know...
Richard Hipp's recent talk at Software Should Work explains that AI agents have been testing SQLite and they've gotten a deluge of new bug reports from the fuzz-like testing they can do. But they do not do this in house; hobbyists and other organizations do this in their own internal agent-driven fuzzing.
huh. ran into almost-certainly this, but blamed it on litestream and rearchitected a bit as a result. will have to see if I can reproduce the issue as we were using with the patched sqlite
I see tailscale also use the pure Go SQLite conversion so I hope this fix will land there soon, it's rapidly become one of my favorite packages for self-contained tools.
Such a good write up . Having explored a little bit of sqlite internals for a codecrafters challenge i was mildly happy i could follow along what was happening .
Everybody here knows that SQLite is not production-ready and does not scale.
This is just proof that it can't be used in real-world applications.
SCNR
I wonder if this also affected litestream disproportionately, because litestream also inserts itself into the checkpoint process.
I would assume they actually moved off of litestream here no? otherwise how can their frequent manual checkpointing even succeed when litestream locks for the same behavior.
I meant does the rare SQLite bug tend to manifest in litestream for the same reason it showed up in Tailscale.
Awesome write up. Finding these bugs in such a well used piece of software is like donating to humanity
> This investigation is a useful reminder: running boring technology in a non-standard way is a risk. The common paths and standard configurations are incredibly well-tested and reliable. Most people use SQLite in a standard configuration and never face this sort of issue. Everything we were doing was a public, documented, supported configuration—but by taking manual control of the checkpointing process and running at our own aggressive pace, we stepped off the well-trodden operational path.
I feel like they missed a key takeaway from their own argument here, they should not be running a non-standard configuration :)
We have very good reasons for our checkpointing model, related to our backup + disaster recover strategy, along with resource cost. It might be worth writing about one day, so I'll not give away all the details, but in very short form, we organize a backup strategy that has minimal pause time, avoids doubling the page cache cost of the database, and enables extremely fast byte-copy restores in disaster recovery.
okay but explain why you are using sqlite and copying the file to S3 instead of using any client/server DB and its online backup feature?
2 replies →
This reminds me why RFC 2119 includes the description it has regarding SHOULD.
Like this is a great practical takeaway, sure, but eh.
Wild to have worked in the industry long enough that 16 years doesn’t feel that long ago.
> Whenever corruption occurred, we had to stop the control plane process on the shard while we repaired or restored the database. This was painful for tailnets on that shard, because their entire control plane disappeared during that recovery window.
Gotta love single points of failure...
This is maybe one of the purest examples of the Bell Curve Meme in software engineering. The things you would do to the system Tailscale operates to eliminate all single points of failure (generally, and in the specific case where, where the "single point of failure" applies only to a small cohort of customers) would make the system less resilient, and increase failures.
Generally, you do complex distributed systems without on-paper single-points-of-failure anywhere when you absolutely have to, because those systems don't have transient failures. That's not mesh networks like Tailscale at all.
As always: https://how.complexsystems.fail/
You don’t need the control plane most of the time. I had a zero downtime headscale upgrade because once the nodes negotiate through the control plane they can talk to each other all the time. The data plane is peer to peer.
It’s problematic because you can’t run connections but it doesn’t stop the world.
Well it's SPoF for a shard not the whole thing so in practice that makes this considerably better. Secondly, it's not a datapath SPoF so presumably all tailscaled processes would continue running and exchanging traffic no problem - you just won't be able to launch any new ones.
What are some solutions to avoid database corruption being single points of failure? I can’t think of any off the top of my head. I don’t think people typically consider database corruption to be a kind of failure common enough to design for, unless you have unusual requirements.
The general answer to this is Byzantine consensus, which cryptocurrency blockchains are designed to solve. If your nodes are willing to fail a little more politely (e.g. no lying, immediately crashing, etc) you can use something cheaper like raft/paxos.
But yeah, it's a lot cheaper to build a reliable system than it is to be resilient.
The shard was already a way to make it not a single point of failure.
This is a great example of outages looking different from the perspective of the operator vs the user. Because there's many shards the blast radius of failure is contained to a small subset of users but for those users it's an outage. The way it's designed you can't lose any shards without impacting users. Compare to say Elasticsearch where it's possible to lose nodes and lose shards without the user noticing. One approach isn't universally better than the other.
1 reply →
Similar bug to the one that plagued Codex until 3 months ago.
Do you happen to have a link?
I lost my Android SMS DB way back in the day (20+ years? dang I'm getting old) because the SMS app's fix when opening the DB detected any issue was to delete it and start fresh.
lol good old ostrich algorithm
love the simplicity of this article. Reminds me so much of foundational software engineering. i love databases def not cosmosdb
Show us the $$ numbers! Or an estimate at least.
I wish Microsoft wrote an apology like that.
Excellently-written article. Bravo!
Learned something new today, thanks
I wonder what happens when you give Claude the old version and the logs and ask it to find the bug.
What logs though? They had to make multiple deployments to get the right log traces in the first place.
"Tailscale Traces Database Corruption to 16y/o ..."
Would have been a superior headline
*sarcasm
The title changed...
Now my joke works even worse :(
> Nobody wanted us to spend six months looking for bugs in SQLite
Most companies wouldn't. Instead, they'd fire the weirdo that came up with the idea of using some weird db, and switch to Postgres like God intended.
I'm not saying either is right, this is not a criticism of Tailscale and their approach, paying the Sqlite maintainers to fix a real bug is commendable, but it certainly doesn't inspire confidence in the "Sqlite in production" hype train.
While Sqlite is indeed boring technology for single-user SQL DBs, for traditional CRUD and network services, Postgres seems to be a much better trodden and much safer path.
>"...and then we discovered an unexpected clue.
We wanted a way to restore service that didn’t involve rolling back to the last known-good backup (which would lose a lot of data) or repairing the known-corrupted database (which was potentially risky). To do this, we built a
transaction logging pipeline.
We streamed every SQL statement that modified the database to a separate log file. Because SQLite is a single-writer database with serializable transactions, our transaction history was completely linear and
deterministic.
(This wouldn’t be true in a multi-writer database like Postgres or MySQL.) Replaying those transactions against the latest known-good backup should restore the database to its most recent state, safely bypassing the corruption.
[...] This pipeline worked, but then it did something even better: it gave us a clue.
[...] To understand what was happening during these faulty checkpoints, the SQLite developers created a new debugging tool for the virtual filesystem layer.
[...] To help diagnose our problem, the SQLite developers created a wrapper around the virtual filesystem that
writes additional tracing information and logs
about changes to the database.
[...] After our next corruption incident, the additional logs from the new tmstmpvfs shim allowed the SQLite developers to find and fix the bug:
a rare data race
in the SQLite source code between a checkpoint and a write transaction."
Great article!
Software Engineering lessons (that repeat in this article!): So called "Heisenbugs" (bugs that make it past developer test harnesses and a company's Quality Assurance (QA) team) that show up post-deployment intermittently and can't be reproduced locally, occur because one or more of the following factors:
1) The lack of Determinism in a software process or processes.
2) The lack of appropriate logging.
3) The lack of the ability to replay a software process, step by exact step, state by exact state, as it has occurred in the field (occurs as an effect of #1 and/or #2).
4) Multi-threaded code; i.e., multiple threads giving rise to race conditions or other very specific intermittent combinatorial/permutational conditions caused by multiple threads and specific sections of code, which due to very large numbers of permutational timing possibilities, were not or could not be exactly tested for in development...
Anyway, great article! A must-read for any Sr. Software Engineer, or any developer that wrestles with hard-to-find-and-fix bugs in the field...
[flagged]
[flagged]
[flagged]
[flagged]
[flagged]
[dead]
[flagged]
[flagged]
[dead]
[dead]
[dead]
[flagged]
you gotta admire the power of using json/b and simple KV stores.
so many people sleep on that.
While technically true as written, it seems to downplay the significance:
> The bug is a data race with tight timing constraints. It is unlikely to occur in common use.
A large customer did experience this corruption, so it's important for people with tailscale's setup update immediately.
> The developers have never been able to reproduce the bug organically and had to add special testing logic to SQLite that deliberately triggers the circumstances of the the bug in order to verify that the issue has been fixed.
Odd not to highlight the sentence where they answer the obvious question "Why Tailscale in particular?":
> They also explained why we were more likely to hit the bug than other SQLite users: we take manual control of the checkpointing process, and we checkpoint very aggressively. Even a bug triggered by a rare condition was bound to hit us eventually.
The bug requires two checkpoints in very quick succession, which presumably isn't something sqlite would do on its own, as it would be a pointless waste of performance.
Tailscale isn't mentioned in the SQLite changelog.
Quick note that data corruption bugs that are impossible to reproduce are not uncommon (perhaps they're the norm). So some amount of head scratching trying to figure out a plausible scenario by which the system could get into the state represented by the smoking remains is often required. Then you attempt to force it into the supposed bad state by modifying code paths accordingly. So the approach used in this case is clever, but it's not particularly unusual in the world of data stores.
I agree it's tricky, but Tailscale had downtime by this bug and the SQLite Changelog downplayed it.
2 replies →
This was a great technical writeup and very interesting to read, but it's not clear to me why once the suspected source of the bug was identified, they seemingly didn't build a automated way to trigger the condition? It seems like that could have cut down on the uncertainty of whether the fix worked over a painfully long period of time.
The sqlite dev team did. It's in the article.
> It could exist that long because it was rare—so rare, the SQLite developers had to add code to deliberately trigger it in their testing environments.
Yes of course I mean the tailscale team. Having an independent way to repro the bug besides waiting for it to happen in prod seems pretty basic eng best practice.