5 ms·

MySql fixes autoincrement bug after 14 years of users complaining about dataloss

42 points by TekMol 9y ago

kanwisher 9y ago
This is quite scary, I wonder how many systems I’ve been on that have had minor unnoticed data corruptions because of this. This is truely a weird thing to not fix
- icedchai 9y ago
  Probably because it's not that scary. In practice, it would generally be a non-issue. 1) mysql server restarts are rare. you'd need a delete followed by a restart to trigger this. 2) you may be using soft deletes, which wouldn't trigger this problem anyway. 3) even if an ID gets reused, it may not actually matter (application dependent.)
  - takeda 9y ago
    According to the comments it actually is a huge issue and multiple people got their data corrupted.
    - icedchai 9y ago
      I've looked at the comments and disagree. For a 15 year old bug, there is actually not that much discussion.
      - takeda 9y ago
        Still, MySQL as any other database has one job, which is keeping data safe. Any issue causing data loss or worse silent data corruption should be considered a critical bug.
        noncoml 9y ago
        There is no data loss caused directly by this bug.
        takeda 9y ago
        Indirect data loss is still a data loss. MongoDB gets a lot of bad rep because of it, MySQL supposed to be on a higher standard.
        icedchai 9y ago
        I'm guessing you didn't use MySQL in the late 90's or early 2000's? It was basically the Mongo of its day, before InnoDB and strict SQL mode..
        takeda 9y ago
        I actually used it, but at the time I didn't know better. I stopped using it when the default engine was still MyISAM, I heard that InnoDB improved it and made it a real database. Now as you mentioned MongoDB it looks like indeed is a sister DB of MySQL. 1. Written by person (people) who did not know anything about databases and learned about them while working on them. 2. Nearly exclusive choice for hip language at the time (PHP for MySQL, NodeJS for MongoDB) 3. Performed much faster than competitive database only to turn out that it was at the cost of durability 4. New engine that supposed to improve durability (WiredTiger vs InnoDB), but even then, it still is plagued by issues caused by initial bad design.
        noncoml 9y ago
        > MongoDB gets a lot of bad rep because of it What are the claims for mongodb?
marindez 9y ago
Now for #20786 https://www.youtube.com/watch?v=oAiVsbXVP6k https://www.youtube.com/watch?v=oAiVsbXVP6k
jlgaddis 9y ago
I'd wager that this only got fixed now because some large Oracle customer got bit hard by this bug.
PhantomGremlin 9y ago
Best comment from the bug report: So we decides to switch on PostreSQL for production... And 14 1/2 years after the first bug report the bug is fixed and: Thank you for the bug report.
andrewstuart 9y ago
They must be busy over there.
noncoml 9y ago
The title is a bit clickbaity. The bug doesn’t cause data loss by itself.
bkm 9y ago
I wonder when it was fixed in Percona's MySQL version, as the reporter of this bug founded Percona in the meantime.
jar3624 9y ago
curious how they did a shutdown of the db. are they saying in all scenarios this causes an the issue ?

42 points by TekMol 9y ago

This is quite scary, I wonder how many systems I’ve been on that have had minor unnoticed data corruptions because of this. This is truely a weird thing to not fix

icedchai 9y ago
Probably because it's not that scary. In practice, it would generally be a non-issue. 1) mysql server restarts are rare. you'd need a delete followed by a restart to trigger this. 2) you may be using soft deletes, which wouldn't trigger this problem anyway. 3) even if an ID gets reused, it may not actually matter (application dependent.)
- takeda 9y ago
  According to the comments it actually is a huge issue and multiple people got their data corrupted.
  - icedchai 9y ago
    I've looked at the comments and disagree. For a 15 year old bug, there is actually not that much discussion.
    - takeda 9y ago
      Still, MySQL as any other database has one job, which is keeping data safe. Any issue causing data loss or worse silent data corruption should be considered a critical bug.
      - noncoml 9y ago
        There is no data loss caused directly by this bug.
        takeda 9y ago
        Indirect data loss is still a data loss. MongoDB gets a lot of bad rep because of it, MySQL supposed to be on a higher standard.
        icedchai 9y ago
        I'm guessing you didn't use MySQL in the late 90's or early 2000's? It was basically the Mongo of its day, before InnoDB and strict SQL mode..
        takeda 9y ago
        I actually used it, but at the time I didn't know better. I stopped using it when the default engine was still MyISAM, I heard that InnoDB improved it and made it a real database. Now as you mentioned MongoDB it looks like indeed is a sister DB of MySQL. 1. Written by person (people) who did not know anything about databases and learned about them while working on them. 2. Nearly exclusive choice for hip language at the time (PHP for MySQL, NodeJS for MongoDB) 3. Performed much faster than competitive database only to turn out that it was at the cost of durability 4. New engine that supposed to improve durability (WiredTiger vs InnoDB), but even then, it still is plagued by issues caused by initial bad design.
        noncoml 9y ago
        > MongoDB gets a lot of bad rep because of it What are the claims for mongodb?

marindez 9y ago

Now for #20786 https://www.youtube.com/watch?v=oAiVsbXVP6k https://www.youtube.com/watch?v=oAiVsbXVP6k

jlgaddis 9y ago

I'd wager that this only got fixed now because some large Oracle customer got bit hard by this bug.

PhantomGremlin 9y ago

Best comment from the bug report: So we decides to switch on PostreSQL for production... And 14 1/2 years after the first bug report the bug is fixed and: Thank you for the bug report.

andrewstuart 9y ago

They must be busy over there.

noncoml 9y ago

The title is a bit clickbaity. The bug doesn’t cause data loss by itself.

bkm 9y ago

I wonder when it was fixed in Percona's MySQL version, as the reporter of this bug founded Percona in the meantime.

jar3624 9y ago

curious how they did a shutdown of the db. are they saying in all scenarios this causes an the issue ?