Re: intermittent long "log file sync" waits

From: Chris Stephens <cstephens16_at_gmail.com>
Date: Thu, 6 Feb 2020 16:19:29 -0600
Message-ID: <CAEFL0szs0BGO0NcM4ZGMbaUfb7Yhkz4T9QQ-YnAdWXHMuOT5MA_at_mail.gmail.com>

I received this update from Oracle:

Discussed this issue with a colleague who mentioned the following--- Bug 25962858 : LOG FILE SYNCS ON A RELATIVELY IDLE DATABASE relates to DENTRY
Closed as not a bug.The solution in that bug was to periodically clear the DENTRY cache.To check the size issueslabtop --onceThis will show something like thisscorreia-linux::oracle> slabtop --once

Active / Total Objects (% used) : 10136104 / 10227087 (99.1%)
Active / Total Slabs (% used) : 273801 / 273801 (100.0%)
Active / Total Caches (% used) : 78 / 108 (72.2%)
Active / Total Size (% used) : 1572176.58K / 1601572.34K (98.2%)

Minimum / Average / Maximum Object : 0.01K / 0.16K / 15.19KOBJS ACTIVE USE OBJ SIZE SLABS OBJ/SLAB CACHE SIZE NAME 5516472 5516241 99% 0.10K 141448 39 565792K buffer_head 1422582 1421855 99% 0.19K 67742 21 270968K dentry

<<<<----------------------------

481664 474527 98% 0.03K 3763 128 15052K kmalloc-32To clear the dentry cache you can issue.... a reboot would do the sameecho 2 > /proc/sys/vm/drop_caches

On Thu, Feb 6, 2020 at 1:31 PM Chris Stephens <cstephens16_at_gmail.com> wrote:

> jeebus. one more time to try and get past moderation.
>
> On Thu, Feb 6, 2020 at 12:55 PM Chris Stephens <cstephens16_at_gmail.com>
> wrote:
>
>> At Sayan's excellent suggestion here is the full snapper output from
>> foreground sessions and lg* processes:
>>
>> https://gist.github.com/cs2018ncsa/e9b0d442247f1434ceef21124bef18c3
>>
>>
>>>>>>>>>
>>>>>>>>>

--
http://www.freelists.org/webpage/oracle-l

Received on Thu Feb 06 2020 - 23:19:29 CET