Should MariaDB touch my.cnf in %post?

Reindl Harald h.reindl at thelounge.net
Sat Feb 16 00:04:47 UTC 2013



Am 15.02.2013 22:32, schrieb Honza Horak:
> On 02/15/2013 3:31 PM, "Jóhann B. Guðmundsson" wrote:
>> On 02/15/2013 01:50 PM, Honza Horak wrote:
>>> I still see the differences between MariaDB and MySQL to be very
>>> little.
>>
>> Difference never the less and difference in behavior ( which mariadbs
>> own benchmarking on their website proof ) which means every server
>> tweak
>> that has been done for the mysql host has to be redone to take
>> whatever
>> changes and features mariadb introduces.
> 
> Can you be more specific, what tweaks you mean?

the difference between theory and life
have fun if something like below hits you after
upgrade and you have not the knowledge i had to
fugure out what happens and fix the app temporary

i have seen queries going 1000 times faster after a
minor update because one bugfix introduced create temp
tables for a simple query on a 16 GB table where the next
bugfix corrected this and made a whle application useable
again - the temporary fix was downgrade or change queries
in dbmail to force a index

https://bugzilla.redhat.com/show_bug.cgi?id=515056

--- dbmail-2.2.11/dbmail-message.c	2009-02-02 15:21:55.000000000 +0100
+++ dbmail-2.2.11-patched/dbmail-message.c	2009-08-01 16:47:38.650404527 +0200
@@ -741,9 +741,10 @@
 static struct DbmailMessage * _fetch_full(struct DbmailMessage *self)
 {
 	char *query_template = "SELECT messageblk, is_header "
-		"FROM %smessageblks "
-		"WHERE physmessage_id = %llu "
-		"ORDER BY messageblk_idnr";
+  "FROM %smessageblks "
+  "FORCE INDEX (physmessage_id_index) "
+  "WHERE physmessage_id = %llu "
+  "ORDER BY messageblk_idnr";
 	return _retrieve(self, query_template);
 }



-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 263 bytes
Desc: OpenPGP digital signature
URL: <http://lists.fedoraproject.org/pipermail/devel/attachments/20130216/104ea809/attachment-0001.sig>


More information about the devel mailing list