Re: intermittent very high waits in LGWR on Linux?

From: bugbear <bugbear_at_trim_papermule.co.uk_trim>
Date: Mon, 04 Jul 2005 10:32:41 +0100
Message-ID: <42c9023d$0$41936$ed2619ec@ptn-nntp-reader03.plus.net>

Noons wrote:

> Fabrizio wrote:
> 
> 

>>I'm only guessing but I would bet sem_op is negative in that syscall
>>(and a portion of pseudo-code on metalink is enforcing by belief).
>>
>>I'd say semop is waiting for a semaphore to be released so I'd look
>>inside oracle more than at the OS level.
>>
>>If you are unsure try simply to change the kernel with a newer 2.4
>>(since you are there) vanilla and see what happens.
>>
>>This call is also interesting:
>>
>>setitimer(ITIMER_REAL, {it_interval={0, 0}, it_value={1, 380000}}, NULL) = 0
>>
>>Your semop will never wait "too long"...
>>

> 
> 
> I've got alarm bells going on about this "1 second" wait:
> *Seen* it mentioned somewhere else, although I'll be
> buggered if I can recall where!
> Searches on Metaclick seem to indicate problems with
> log sync waits at early 9ir2 releases: 9.2.0.1 and such.
> There is nothing remotely like that for 9.2.0.6.  But
> I don't know which particular level Bugsy is using.
> 

1 last possible diagnostic fact. If the DB hasn't seen any activity for "a long time" (more than 20 minutes, at least), the first test run goes very quickly. I just got 100 insertions a second, against my normal "worst" of 8.

    BugBear Received on Mon Jul 04 2005 - 04:32:41 CDT