xfs
[Top] [All Lists]

Re: nginx blocked task when XFS partition is very full

To: xfs@xxxxxxxxxxx
Subject: Re: nginx blocked task when XFS partition is very full
From: Stan Hoeppner <stan@xxxxxxxxxxxxxxxxx>
Date: Fri, 02 Sep 2011 08:02:05 -0500
In-reply-to: <3972BA0F-0E95-4758-9DDB-BC6197621C87@xxxxxxxxxxx>
References: <3972BA0F-0E95-4758-9DDB-BC6197621C87@xxxxxxxxxxx>
User-agent: Mozilla/5.0 (Windows NT 5.1; rv:6.0.1) Gecko/20110830 Thunderbird/6.0.1
On 9/2/2011 7:23 AM, Sean Purdy wrote:
Hi,


This is on 2.6.32-32-server from Ubuntu Lucid 10.04 LTS

We have nginx fileserving from four disks on a machine.  We're using mogilefs as a 
distributed file system.  We've come across an issue where when the filesystem is very 
full, and nginx tries to write to it, we get "task blocked for more than 120 
seconds" messages from the kernel, nginx hangs and we can no longer access the fs.  
When a file is uploaded via nginx, nginx writes to a .tmp directory on the disk before 
moving the file to its destination.  nginx can't be kill -9ed, it needs a reboot.

Has this sort of issue been fixed since 2.6.32?

It would help to see df output for space and inodes of the XFS filesystem in question.

--
Stan

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