Comments on: Can you Trust CHECK TABLE ? http://www.mysqlperformanceblog.com/2008/09/11/can-you-trust-check-table/ Everything about MySQL Performance Sat, 07 Nov 2009 18:35:44 -0800 http://wordpress.org/?v=2.8.4 hourly 1 By: peter http://www.mysqlperformanceblog.com/2008/09/11/can-you-trust-check-table/comment-page-1/#comment-354642 peter Fri, 12 Sep 2008 17:29:33 +0000 http://www.mysqlperformanceblog.com/?p=497#comment-354642 Pat, Actually there IS a second copy of meta data in MYI files.... which can be checked against what is in .frm files but it is not. This is artificial example of course but believe me I've seen mismatching .frm files from MYI/MYD files many times. In particular for MyISAM - because this is table type which is advertised as being possible to move around... so people do it. Not always correctly. The example with replication is wrong. If you delete the binary log file from the system replication will complain just as you would expect. If you trash binary log it may not... though it is common complain too what checksums are missing in binary logs. Pat,

Actually there IS a second copy of meta data in MYI files…. which can be checked against what is in .frm files but it is not. This is artificial example of course but believe me I’ve seen mismatching .frm files from MYI/MYD files many times. In particular for MyISAM – because this is table type which is advertised as being possible to move around… so people do it. Not always correctly.

The example with replication is wrong. If you delete the binary log file from the system replication will complain just as you would expect. If you trash binary log it may not… though it is common complain too what checksums are missing in binary logs.

]]> By: James Aylett http://www.mysqlperformanceblog.com/2008/09/11/can-you-trust-check-table/comment-page-1/#comment-354638 James Aylett Fri, 12 Sep 2008 17:09:33 +0000 http://www.mysqlperformanceblog.com/?p=497#comment-354638 You can detect this situation to some extent by having some kind of hash of the frm in the myd/myi files. On the other hand, you can always break systems by messing directly with their internal files, so I wouldn't argue that this is as important as perhaps providing tools for managing rollback/restore of these files as a group. Or just telling you not to do that sort of thing :-) You can detect this situation to some extent by having some kind of hash of the frm in the myd/myi files. On the other hand, you can always break systems by messing directly with their internal files, so I wouldn’t argue that this is as important as perhaps providing tools for managing rollback/restore of these files as a group. Or just telling you not to do that sort of thing :-)

]]> By: Pat http://www.mysqlperformanceblog.com/2008/09/11/can-you-trust-check-table/comment-page-1/#comment-354608 Pat Fri, 12 Sep 2008 16:18:16 +0000 http://www.mysqlperformanceblog.com/?p=497#comment-354608 How's check table even going to if there's an error here w/o carrying a second copy of the metadata in the MYD file? Seems like he can only reasonably be expected to compare the data in the MYD with the data in the FRM and if they look consistent all is well. If you spoof the metadata like this, I think that qualifies as an event the product can't reasonably be expected to recover from. Like going out and deleting the binary log from the file system and then opening a bug that replication is broken. How’s check table even going to if there’s an error here w/o carrying a second copy of the metadata in the MYD file? Seems like he can only reasonably be expected to compare the data in the MYD with the data in the FRM and if they look consistent all is well.

If you spoof the metadata like this, I think that qualifies as an event the product can’t reasonably be expected to recover from. Like going out and deleting the binary log from the file system and then opening a bug that replication is broken.

]]> By: peter http://www.mysqlperformanceblog.com/2008/09/11/can-you-trust-check-table/comment-page-1/#comment-354535 peter Fri, 12 Sep 2008 06:59:39 +0000 http://www.mysqlperformanceblog.com/?p=497#comment-354535 Harmut, I remember I ran into this 4-5 years ago when I was still with MySQL and it was described as product feature. Though now it is the time to fix it I think - especially as there will be more and more meta data which will need to be tested. For example Global foreign keys with MyISAM - how do you check they are in order if there is no such command. What if you add CHECK constrains ? Harmut,

I remember I ran into this 4-5 years ago when I was still with MySQL and it was described as product feature.
Though now it is the time to fix it I think – especially as there will be more and more meta data which will need to be tested. For example Global foreign keys with MyISAM – how do you check they are in order if there is no such command. What if you add CHECK constrains ?

]]> By: hartmut http://www.mysqlperformanceblog.com/2008/09/11/can-you-trust-check-table/comment-page-1/#comment-354533 hartmut Fri, 12 Sep 2008 06:20:50 +0000 http://www.mysqlperformanceblog.com/?p=497#comment-354533 Hi Peter, is there a bug report for this already or should we create one? Hi Peter,

is there a bug report for this already or should we create one?

]]>