xfs
[Top] [All Lists]

Re: xlog_write: reservation ran out. Need to up reservation

To: xfs@xxxxxxxxxxx
Subject: Re: xlog_write: reservation ran out. Need to up reservation
From: Thomas Klaube <thomas@xxxxxxxxxx>
Date: Thu, 21 Aug 2014 09:24:14 +0200 (CEST)
Cc: Dave Chinner <david@xxxxxxxxxxxxx>
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <20140819225507.GS20518@dastard>
References: <159192779.3859815.1408461799560.JavaMail.zimbra@xxxxxxxxxx> <362338960.3862279.1408462470243.JavaMail.zimbra@xxxxxxxxxx> <20140819225507.GS20518@dastard>
Thread-index: /XT7UEmjHs8oyGkE0oOjhoQbFCeS8g==
Thread-topic: xlog_write: reservation ran out. Need to up reservation
----- UrsprÃngliche Mail -----
> Von: "Dave Chinner" <david@xxxxxxxxxxxxx>
> An: "Thomas Klaube" <thomas@xxxxxxxxxx>
> CC: xfs@xxxxxxxxxxx
> Gesendet: Mittwoch, 20. August 2014 00:55:07
> Betreff: Re: xlog_write: reservation ran out. Need to up reservation

Hi,

> Can you post the fio job configuration?

first I run this job for 600 sec:
wtk@ubuntu ~ $ cat write.fio
[rnd]
rw=randwrite
ramp_time=30
runtime=600
time_based
gtod_reduce=1
size=100g
refill_buffers=1
directory=.
iodepth=64
direct=1
blocksize=16k
numjobs=64
nrfiles=1
group_reporting
ioengine=libaio
loops=1

Then I run this job for 2hours:
wtk@ubuntu ~ $ cat random.fio
[rnd]
rw=randrw
ramp_time=30
runtime=7200
time_based
rwmixread=30
size=100g
refill_buffers=1
directory=.
iodepth=64
direct=1
blocksize=4k
numjobs=64
group_reporting
ioengine=libaio
loops=1

I run this workload on 2 devices in parallel. One is the bcache device (with 
xfs), the other is
a non cached device. The random.fio job causes the problem on the bcache device 
after ~30-75mins.

> http://xfs.org/index.php/XFS_FAQ#Q:_What_information_should_I_include_when_reporting_a_problem.3F

I have sent a mail with all collected data to Dave.

> Run the workload directly on the SSD rather than with bcache. Use
> mkfs parameters to give you 8 ags and the same size log, and see
> if you get the same problem.

I created a xfs direclty on the SSD:
mkfs.xfs -f -d agcount=8 -l size=521728b /dev/sdc1

Then I started tho fio jobs as described above for 10 hours. I could not 
reproduce the
problem. I will send a mail to the bcache mailing list as well...

Thanx and Regards
Thomas

<Prev in Thread] Current Thread [Next in Thread>