https://jira.whamcloud.com This file is an XML representation of an issue en-us 9.4.14 940014 05-12-2023 https://jira.whamcloud.com/browse/LU-7733 Lustre <p>We would like to see a higher time resolution in ptlrpc_lprocfs_svc_req_history_show output.</p> <p>Here we see it is only printing up to a second.</p> <div class="code panel" style="border-width: 1px;"><div class="codeContent panelContent"> <pre class="code-java">seq_printf(s, LPD64<span class="code-quote">":%s:%s:x"</span>LPU64<span class="code-quote">":%d:%s:%ld:%lds(%+lds) "</span>, req-&gt;rq_history_seq, libcfs_nid2str(req-&gt;rq_self), libcfs_id2str(req-&gt;rq_peer), req-&gt;rq_xid, req-&gt;rq_reqlen, ptlrpc_rqphase2str(req), req-&gt;rq_arrival_time.tv_sec, req-&gt;rq_sent - req-&gt;rq_arrival_time.tv_sec, req-&gt;rq_sent - req-&gt;rq_deadline); </pre> </div></div> LU-7733

ptlrpc_lprocfs_svc_req_history_show (Seq_printf) time resolution should be higher than 1 second.

Improvement Major Resolved Fixed Bob Glossman Mahmoud Hanafi Tue, 2 Feb 2016 20:47:17 +0000 Mon, 19 Sep 2016 16:15:03 +0000 Mon, 16 May 2016 21:42:59 +0000 Lustre 2.5.3 Lustre 2.9.0 0 7 <p>its trivial to add extra resolution, the biggest question is that of compatibility. While we d o not have any tools that parse it, there might be something in the community that is not expecting anything but an integer there.</p> <p>Perhaps we need to ask around in lustre-devel?</p> <p>Bob</p> <p>Can you push a provisional change here while we assess the impact of making this change?</p> <p>Thanks</p> <p>Peter</p> <p>not at all clear what is wanted here. I can easily print a higher resolution value in place of req-&gt;rq_arrival_time.tv_sec, just convert to jiffies or microseconds or something. not very human readable compared to seconds IMHO. don't know what to do about the diff values, req-&gt;rq_sent - req-&gt;rq_arrival_time.tv_sec and req-&gt;rq_sent - req-&gt;rq_deadline. rq_sent is in seconds so I think only seconds diffs are easy to come by.</p> <p>We would be interested at lest microseconds. In the rpctrace debug dump the completion times are in micro-seconds. </p> <p>Bob Glossman (bob.glossman@intel.com) uploaded a new patch: <a href="http://review.whamcloud.com/18335" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/18335</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-7733" title="ptlrpc_lprocfs_svc_req_history_show (Seq_printf) time resolution should be higher than 1 second." class="issue-link" data-issue-key="LU-7733"><del>LU-7733</del></a> ptlrpc: print times in microseconds<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: 1<br/> Commit: 4343fc1738f22b24042b12616b5ba16567ff30fe</p> <p>still not sure I'm really addressing the request exactly, but have pushed a mod with my best guess. it's at least a starting point for discussion.</p> <p>I agree with Andreas comment, what we want is seconds.microsecond.</p> <p>I did try out the patch. One the MDS it is working but not the output we were looking for...</p> <div class="code panel" style="border-width: 1px;"><div class="codeContent panelContent"> <pre class="code-java">6250131278572683267:10.151.26.117@o2ib:12345-10.151.26.120@o2ib:x1525908159598524:224:Complete:1455222088536099:-536099us(-192000000us) opc 400 6250131278572683271:10.151.26.117@o2ib:12345-10.151.26.120@o2ib:x1525908159598504:224:Complete:1455222088536106:-536106us(-192000000us) opc 400 6250127494744440832:10.151.26.117@o2ib:12345-10.151.27.19@o2ib:x1525733058816312:264:Complete:1455221207545364:-545364us(-192000000us) opc 700 </pre> </div></div> <p>There is something wrong on the OSS as it doesn't print anything.</p> <div class="code panel" style="border-width: 1px;"><div class="codeContent panelContent"> <pre class="code-java">nbp1-oss1 /proc/fs/lustre/ost/OSS/ost_io # cat req_buffer_history_max 5000 nbp1-oss1 /proc/fs/lustre/ost/OSS/ost_io # cat req_buffer_history_len 1250 nbp1-oss1 /proc/fs/lustre/ost/OSS/ost_io # cat req_history nbp1-oss1 /proc/fs/lustre/ost/OSS/ost_io # </pre> </div></div> <p>I don't understand the negative numbers here....</p> <div class="code panel" style="border-width: 1px;"><div class="codeContent panelContent"> <pre class="code-java">6266059553153941504:10.151.27.60@o2ib:12345-10.151.7.174@o2ib:x1529670166131456:328:Complete:1458930678.979710:-1.020290s(-192.0s) 6266059553091026945:10.151.27.60@o2ib:12345-10.151.53.147@o2ib:x1529728924728644:328:Complete:1458930678.964339:-1.035661s(-192.0s) 6266059553440268289:10.151.27.60@o2ib:12345-10.151.54.82@o2ib:x1529750902839948:328:Complete:1458930679.001036:-1.998964s(-192.0s) 6266059553444003841:10.151.27.60@o2ib:12345-10.151.54.82@o2ib:x1529750902839960:328:Complete:1458930679.001944:-1.998056s(-192.0s) 6266059553454292994:10.151.27.60@o2ib:12345-10.151.40.144@o2ib:x1529023858938888:328:Complete:1458930679.004457:-1.995543s(-192.0s) 6266059553458880514:10.151.27.60@o2ib:12345-10.151.40.144@o2ib:x1529023858938904:328:Complete:1458930679.005578:-1.994422s(-192.0s) 6266059553481883650:10.151.27.60@o2ib:12345-10.151.40.144@o2ib:x1529023858938924:328:Complete:1458930679.011186:-1.988814s(-192.0s) 6266059553533132803:10.151.27.60@o2ib:12345-10.153.10.116@o2ib233:x1529703480218204:328:Complete:1458930679.023701:-1.976299s(-192.0s) </pre> </div></div> <p>the negative numbers are because the values shown are deltas, not absolute times. if I'm not mistaken the values shown without the patch were always 0 or negative too.</p> <p>Bob, the review #18335 has conflict with b2_7_fe. 'Git blame' points to <a href="https://jira.whamcloud.com/browse/LU-6070" title="provide separate buffers for libcfs_*2str()" class="issue-link" data-issue-key="LU-6070"><del>LU-6070</del></a>.<br/> Do we need <a href="https://jira.whamcloud.com/browse/LU-6070" title="provide separate buffers for libcfs_*2str()" class="issue-link" data-issue-key="LU-6070"><del>LU-6070</del></a> review #907a321c? Or can you provide a b2_7_fe back port?</p> <p>Jay,<br/> There is no current plan to back port to b2_7_fe. It can probably be done if thought important. It's a priority call, I'm not the decider.</p> <p>Hi Bob, I can do the back port myself, given that Intel tells me that this patch would still work as expected without <a href="https://jira.whamcloud.com/browse/LU-6070" title="provide separate buffers for libcfs_*2str()" class="issue-link" data-issue-key="LU-6070"><del>LU-6070</del></a>.</p> <p>Jay,<br/> As far as I can tell <a href="https://jira.whamcloud.com/browse/LU-6070" title="provide separate buffers for libcfs_*2str()" class="issue-link" data-issue-key="LU-6070"><del>LU-6070</del></a> changes have nothing whatsoever to do with this mod. changes from both happen to overlap.</p> <p>I can't promise 100% that <a href="https://jira.whamcloud.com/browse/LU-6070" title="provide separate buffers for libcfs_*2str()" class="issue-link" data-issue-key="LU-6070"><del>LU-6070</del></a> will never be going into b2_7_fe too, but I don't think it's a requirement for making changes from this mod apply there.</p> <p>We don't usually port to branches until after a mod lands in master. This one hasn't. Hasn't even gotten any feedback from the originator of the request if it meets his needs.</p> <p>Oops, sorry, Bob. I did not read the contents of this ticket. Yes the mod was simply to increase the time resolution in print messages. Actually I already cherry-picked the mod to b2_5_fe before. <img class="emoticon" src="https://jira.whamcloud.com/images/icons/emoticons/wink.png" height="16" width="16" align="absmiddle" alt="" border="0"/></p> <p>Oleg Drokin (oleg.drokin@intel.com) merged in patch <a href="http://review.whamcloud.com/18335/" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/18335/</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-7733" title="ptlrpc_lprocfs_svc_req_history_show (Seq_printf) time resolution should be higher than 1 second." class="issue-link" data-issue-key="LU-7733"><del>LU-7733</del></a> ptlrpc: print times in microseconds<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: <br/> Commit: 1929bfe81379a9c986bfa4202265f76d201a49bc</p> <p>Landed for 2.9</p> Related Development Rank 1|hzy01j: Rank (Obsolete) 9223372036854775807