xfs
[Top] [All Lists]

Re: Still seeing hangs in xlog_grant_log_space

To: Dave Chinner <david@xxxxxxxxxxxxx>
Subject: Re: Still seeing hangs in xlog_grant_log_space
From: Mark Tinguely <tinguely@xxxxxxx>
Date: Wed, 09 May 2012 21:10:05 -0500
Cc: Juerg Haefliger <juergh@xxxxxxxxx>, xfs@xxxxxxxxxxx
In-reply-to: <4FAADCA8.6030702@xxxxxxx>
References: <20120425223845.GX9541@dastard> <CADLDEKvYkpUnMrqdMyqCmsYrZcUtiJ6ZRhrRu_ERTjn=r7M3Pg@xxxxxxxxxxxxxx> <20120426224412.GA9541@dastard> <CADLDEKs6oMDA-6OhmcFxyRoBVpduKtSput=53TQGn9NCAOXC1Q@xxxxxxxxxxxxxx> <20120426230738.GB9541@dastard> <CADLDEKuKLeYiqhQW0E9g_bS0VXoxPGPOck3N004Pxg4_Opbzow@xxxxxxxxxxxxxx> <20120427110922.GF9541@dastard> <CADLDEKtUHAGcOPT1jtcvyJVk+zsoL5_thYFtHJYs+w=6EGuVSA@xxxxxxxxxxxxxx> <CADLDEKs4YbNzj2c0HKHwSdUfKy0efdQRe1rOsWDkWUgd+BOGHw@xxxxxxxxxxxxxx> <20120507225944.GA5091@dastard> <20120509073507.GQ5091@dastard> <4FAADCA8.6030702@xxxxxxx>
User-agent: Mozilla/5.0 (X11; FreeBSD amd64; rv:9.0) Gecko/20120122 Thunderbird/9.0
On 05/09/12 16:07, Mark Tinguely wrote:


Maybe I have a corrupted version of his first trace, it looks like there
are 2 series of log grant/write/head/tail sequences. These sequences are
not even close to each other:

188.116687: FSYNC_TS empty empty 1 847894476 1 847894476 1 1655998 1
1655971
188.116939: FSYNC_TS empty empty 1 847894476 1 847894476 1 1655998 1
1655971
188.117755: CREATE empty empty 4440 166388 4440 166388 4440 312 4440 310
188.117784: FSYNC_TS empty empty 1 847894476 1 847894476 1 1655998 1
1655971
188.117902: FSYNC_TS empty empty 1 847894476 1 847894476 1 1655998 1
1655971
188.118249: CREATE empty empty 4440 166844 4440 166844 4440 312 4440 310
188.118350: CREATE empty empty 4440 167300 4440 167300 4440 312 4440 310
188.118628: FSYNC_TS empty empty 4440 167300 4440 167300 4440 312 4440 310
188.118837: FSYNC_TS empty empty 1 847894476 1 847894476 1 1655998 1
1655971
188.118868: FSYNC_TS empty empty 1 847894476 1 847894476 1 1655998 1
1655971

Oops, there are multiple devices in that trace.


I notice in the trace_report_180secs.txt file, lsn (ail_push and the ungrants) displays are not correct. It acts like the BLOCK_LSN() is shifting too much. Block sequence numbers never make it too far above 1100 before incrementing the cycle. I think the problem is the trace not the sequence numbers.

--Mark Tinguely.

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