[Maria-discuss] Fwd: [Dbmail] MariaDB and dbmail
please can someone from the MariaDB developers MariaDB 5.5.32 is running here on test-machines this is todays feedback of the dbmail core-developer and beause more and more distributions including Fedora/OpenSuSE and AFAIK even RHEL7 switch to MariaDB this is a *serious* problem which is not what a "drop-in replacement" should do and a show-stopper rollout Fedora 19 at all in the near future
Anyway, MariaDB doesn't even pass the basic unit-tests
Looks like this maybe libzdb related. The ordering of the query retrieving mime parts is ok on the console, but *not* when looping over the result in libzdb.
MariaDB is *no-go* at this moment. Just what I experienced a year ago or so. I'm sorry, but I don't think I'll hold 3.1.4 for that not a regression at all.
-------- Original-Nachricht --------
Betreff: Re: [Dbmail] MariaDB and dbmail
Datum: Tue, 02 Apr 2013 19:47:15 +0400
Von: Sergej Pupykin
MariaDB is fully compatible with MySQL. Migration is done. So far so good. Phewww!...
Now I need to do some research on tuning mariadb if needed.
I found that after migration following SQL returns wrong sorted data: SELECT l.part_key, l.part_depth, l.part_order, l.is_header, DATE_FORMAT(ph.internal_date,GET_FORMAT(DATETIME,'ISO')), data FROM dbmail_mimeparts p JOIN dbmail_partlists l ON p.id = l.part_id JOIN dbmail_physmessage ph ON ph.id = l.physmessage_id WHERE l.physmessage_id = <ID> ORDER BY l.part_key,l.part_order ASC; however when I run it from command line it works well. Probably it does not fail always because of not all emails damaged. Not sure if it caused by migration to mariadb, but problem appeared near same time.
Hi!
"Reindl" == Reindl Harald
writes:
Reindl> please can someone from the MariaDB developers Reindl> MariaDB 5.5.32 is running here on test-machines Reindl> this is todays feedback of the dbmail core-developer and beause Reindl> more and more distributions including Fedora/OpenSuSE and AFAIK Reindl> even RHEL7 switch to MariaDB this is a *serious* problem which is Reindl> not what a "drop-in replacement" should do and a show-stopper Reindl> rollout Fedora 19 at all in the near future
Anyway, MariaDB doesn't even pass the basic unit-tests
Looks like this maybe libzdb related. The ordering of the query retrieving mime parts is ok on the console, but *not* when looping over the result in libzdb.
MariaDB is *no-go* at this moment. Just what I experienced a year ago or so. I'm sorry, but I don't think I'll hold 3.1.4 for that not a regression at all.
I never saw a final follow up on this, so I thought I would post some observations. This was a bug in that in some cases with Aria temporary tables it was not always guaranteed that one would read rows in the same order they where inserted. This bug has been around a long time in MariaDB and in spite of having lots of MariaDB installations this has never been noticed or reported. In any case, it's now fixed in MariaDB 5.5.33 and also in MariaDB 5.1, 5.2 and 5.3. We are doing our best to ensure that MariaDB is a full MySQL drop in replacement and in most cases it is. As far as I can remember during the last 4 years we have had only a handfull of serious bugs like this. We have also done our best to document the differences between MySQL and MariaDB at: https://mariadb.com/kb/en/mariadb-vs-mysql-compatibility/ In general, my own experience is that it's easier to move to MariaDB than to a newer MySQL version. In both cases there may be new bugs or new behavior. We at MariaDB are doing our best to close everything that affects other open source projects and distributions very fast. Just report the bugs in Jira or here on the email list and we are doing our best to quickly fix it. In any case, sorry about the trouble with the ORDER BY bug. Regards, Monty
Hi! thank you very much - the timeframe report/fix was really impressive best greetings from Vienna! Harry Am 01.10.2013 09:09, schrieb Michael Widenius:
Hi!
"Reindl" == Reindl Harald
writes: Reindl> please can someone from the MariaDB developers Reindl> MariaDB 5.5.32 is running here on test-machines
Reindl> this is todays feedback of the dbmail core-developer and beause Reindl> more and more distributions including Fedora/OpenSuSE and AFAIK Reindl> even RHEL7 switch to MariaDB this is a *serious* problem which is Reindl> not what a "drop-in replacement" should do and a show-stopper Reindl> rollout Fedora 19 at all in the near future
Anyway, MariaDB doesn't even pass the basic unit-tests
Looks like this maybe libzdb related. The ordering of the query retrieving mime parts is ok on the console, but *not* when looping over the result in libzdb.
MariaDB is *no-go* at this moment. Just what I experienced a year ago or so. I'm sorry, but I don't think I'll hold 3.1.4 for that not a regression at all.
I never saw a final follow up on this, so I thought I would post some observations.
This was a bug in that in some cases with Aria temporary tables it was not always guaranteed that one would read rows in the same order they where inserted.
This bug has been around a long time in MariaDB and in spite of having lots of MariaDB installations this has never been noticed or reported.
In any case, it's now fixed in MariaDB 5.5.33 and also in MariaDB 5.1, 5.2 and 5.3.
We are doing our best to ensure that MariaDB is a full MySQL drop in replacement and in most cases it is. As far as I can remember during the last 4 years we have had only a handfull of serious bugs like this.
We have also done our best to document the differences between MySQL and MariaDB at: https://mariadb.com/kb/en/mariadb-vs-mysql-compatibility/
In general, my own experience is that it's easier to move to MariaDB than to a newer MySQL version. In both cases there may be new bugs or new behavior.
We at MariaDB are doing our best to close everything that affects other open source projects and distributions very fast.
Just report the bugs in Jira or here on the email list and we are doing our best to quickly fix it.
In any case, sorry about the trouble with the ORDER BY bug.
Regards, Monty
participants (2)
-
Michael Widenius
-
Reindl Harald