On Fri, Nov 6, 2009 at 3:54 AM, Philip Pokorny
<[email protected]> wrote:
> ======= Update
> At Chris's suggestion, I changed the scst threading to a *single* thread.  I
> also changed the number of outstanding commands to 32 (from 64)  We have a
> relatively large number of initiators (50, currently testing with only 4)
> but each initiator has it's own dedicated, non-shared set of LVM "block"
> devices.  So it does not seem necessary to set the device queue size to
> initiators*device_queue_size.  I set the total device queued commands to 37.
>
> Re-running the test, the storage is "stable" and completed the first phase
> of the test (create and check the filesystem) but now two of four clients
> have "live-locked" waiting for an I/O to complete for the last 15 minutes or
> so.
>
> On the initiators iostat looks like this:
>
> Device:         rrqm/s   wrqm/s   r/s   w/s    rkB/s    wkB/s avgrq-sz
> avgqu-sz   await  svctm  %util
> sdb               0.00     0.00  0.00  0.00     0.00     0.00     0.00
> 160.02    0.00   0.00 100.01
> sdb               0.00     0.00  0.00  0.00     0.00     0.00     0.00
> 160.02    0.00   0.00 100.01
> sdb               0.00     0.00  0.00  0.00     0.00     0.00     0.00
> 160.02    0.00   0.00 100.01
> sdb               0.00     0.00  0.00  0.14     0.00     8.14   114.00
> 159.34 180114.00 7002.00 100.03
> sdb               0.00     0.00  0.00  0.00     0.00     0.00     0.00
> 160.02    0.00   0.00 100.01
> sdb               0.00     0.00  0.00  0.00     0.00     0.00     0.00
> 160.02    0.00   0.00 100.01
>
> This tells me that there is a pending I/O waiting to be completed but it
> seems to have been lost on the server, because this is taking much too
> long.  There are 7 seconds "between" each line of output above so that's
> almost 30 seconds of output with *no* change in the I/O status.
>
> The "gzip | tar -x" I was running is "hung"

Hello Phil,

Can you please post the SCST target logs available for the above scenario ?

Bart.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to