LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: Again: LVS/Apache cluster freezes from time to time

To: "LinuxVirtualServer.org users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: Again: LVS/Apache cluster freezes from time to time
From: Jan Abraham <jan_abraham@xxxxxxx>
Date: Wed, 12 Nov 2003 19:48:40 +0100
> > Problem A:
> > - Poorly written PHP application (lucky me, not my fault...) -> tons of
> > PHP includes on every request
> 
> lots of disk accesses?

Yesss...

> 
> > - Lack of noatime,nodiratime in /etc/fstab
> 
> more disk accesses?

Yesss...

> 
> > - Use of ext3 with the default data mode (ordered)
> 
> what's this about?

I'm not an expert in filesystems, but I can imagine that the ext3
journal ran out of space and holds the system until all entries were
written on their respective places. Just an idea.
"man mount" suggests to use "writeback" as data mode to improve
performance, with the risk of having some files containing old data
fragments after a crash.

The reason I choosed a journaling file system was to minimize down time
after a crash. For now, I've switched back to ext2, but I'll do some new
attempts with the suggested writeback mode on ext3.


> glad you figured it out. I'll put it in the HOWTO sometime as a
> gruesome story that turned out not to be LVS, but for which there
> was no explanation for quite a while.

Yes, I think it shall be written with bright red letters: 
"do not use journaling filesystem without noatime,nodiratime on a high
traffic website".

:o)

Thanks,
Jan

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