|
From: Douglas T. <dou...@gm...> - 2009-01-06 13:18:13
|
On Wed, Dec 31, 2008 at 1:38 PM, Douglas Tosi <dou...@gm...> wrote: > On Wed, Dec 31, 2008 at 12:45 PM, Dmitry Yemanov <fir...@ya...> wrote: >> Douglas Tosi wrote: >>> In the meanwhile, is the above information useful? Unlike the case >>> stated in core-2209 my test does not contain hundreds of attachments, >>> but the server is highly flooded with I/O requests. Is core-2209 >>> supposed to fix this too? >> >> Unlikely, I'm afraid. Anyway, I'd prefer to have a copy of your test as >> soon as you manage to simplify it :-) > > OK. Dmitry, all, build WI-V2.1.2.18112 is crashing here. I really don't know if it is something wrong with Firebird or I'm doing something *really* stupid. Anyway, the test is as follows. 1) Using .net client (both 2.1 and 2.5 beta1) create 1000 attachments. After this step fbserver memory rises from 7MB to 45MB. 2) start another attachment that starts a transaction, executes the following monitoring sql and commits every second: select mon$attachment_id id, mon$state state from mon$attachments mat where mat.mon$attachment_id <> current_connection order by mat.mon$attachment_id 3) each of the previous 1000 attachments start a transaction Firebird server crashes. The log file has nothing except some 10054 messages. You may download the test here: http://www.sinatica.com/downloads/TestCORE-2209.7z - .Net runtime required - Click "connect" and when the message "started" appears click "start transactions" to start steps 2 and 3. I ran this on Vista64 with Firebird 32bit WI-V2.1.2.18112 Firebird 2.1. The test runs fine with 2.1.1. Thank you, -- Douglas Tosi www.sinatica.com |