[Top] [All Lists]

Re: xfs_repair in traversing filesystem

To: Linux XFS <xfs@xxxxxxxxxxx>
Subject: Re: xfs_repair in traversing filesystem
From: pg_xf2@xxxxxxxxxxxxxxxxx (Peter Grandi)
Date: Wed, 2 Mar 2011 19:47:02 +0000
In-reply-to: <AANLkTimJ0tdvSy+7MSykNm4mW1XoO8PQ_-G9UBonFh8V@xxxxxxxxxxxxxx>
References: <AANLkTimJ0tdvSy+7MSykNm4mW1XoO8PQ_-G9UBonFh8V@xxxxxxxxxxxxxx>
> Our 35TB XFS filesystem developed some problems and was
> throwing these: [ ... ]

How many drives support that filesystem?
How many inodes does it have?

> so we're running xfs_repair on this filesystem.  It finds this
> inode and everything looks good, but then it hangs at [ ... ]
> - traversing filesystem ...  previous posts suggest running
> xfs_repair -P -o bhash=1024 [ ... ]

It depends on how fast you expect that to happen and how much
memory you have. 'man xfs_repair' says that "The default size is
set to use up the remainder of 75% of the system's physical RAM
size" which seems pretty good to me. You may want to increase
the '-o ihash' value too, or '-m'. Also check the '-P' option.

BTW I just noticed in the same 'man xfs_repair' the option that
"can significantly reduce repair times on concat based filesystems"
which is so amusing...

<Prev in Thread] Current Thread [Next in Thread>
  • Re: xfs_repair in traversing filesystem, Peter Grandi <=