<html><body bgcolor="#FFFFFF"><p>Hi Tejas,<br><br>Yes, most likely those 4k writes are the HAWC writes...hopefully those 4KB writes have a lower latency than the 4k writes to your data pool so you are realizing the benefits.<br><br>Dean <br><br><br><img width="16" height="16" src="cid:1__=07BB0A91DFE1314A8f9e8a93df938690918c07B@" border="0" alt="Inactive hide details for Tejas Rao ---08/01/2016 01:42:33 PM---I am not 100% sure what the workload of the VMs is. We have 100"><font color="#424282">Tejas Rao ---08/01/2016 01:42:33 PM---I am not 100% sure what the workload of the VMs is. We have 100's of VMs all used differently, so t</font><br><br><font size="2" color="#5F5F5F">From: </font><font size="2">Tejas Rao <raot@bnl.gov></font><br><font size="2" color="#5F5F5F">To: </font><font size="2">gpfsug main discussion list <gpfsug-discuss@spectrumscale.org></font><br><font size="2" color="#5F5F5F">Date: </font><font size="2">08/01/2016 01:42 PM</font><br><font size="2" color="#5F5F5F">Subject: </font><font size="2">Re: [gpfsug-discuss] HAWC (Highly available write cache)</font><br><font size="2" color="#5F5F5F">Sent by: </font><font size="2">gpfsug-discuss-bounces@spectrumscale.org</font><br><hr width="100%" size="2" align="left" noshade style="color:#8091A5; "><br><br><br><font size="4">I am not 100% sure what the workload of the VMs is. We have 100's of VMs all used differently, so the workload is rather mixed. <br><br>I do see 4K writes going to "system" pool, they are tagged as "logData" in 'mmdiag --iohist'. But I also see 4K writes going to the data drives, so it looks like everything is not getting coalesced and these are random writes. <br><br>Could these 4k writes labelled as "logData" be the writes going to HAWC log files?<br><br></font><br><font size="4">On 8/1/2016 15:50, Dean Hildebrand wrote:</font><ul><ul><font size="4">Hi Tejas,<br><br>Do you know the workload in the VM? <br><br>The workload which enters into HAWC may or may not be the same as the workload that eventually goes into the data pool....it all depends on whether the 4KB writes entering HAWC can be coalesced or not. For example, sequential 4KB writes can all be coalesced into a single large chunk. So 4KB writes into HAWC will convert into 8MB writes to data pool (in your system). But random 4KB writes into HAWC may end up being 4KB writes into the data pool if there are no adjoining 4KB writes (i.e., if 4KB blocks are all dispersed, they can't be coalesced). The goal of HAWC though, whether the 4KB blocks are coalesced or not, is to reduce app latency by ensuring that writing the blocks back to the data pool is done in the background. So while 4KB blocks may still be hitting the data pool, hopefully the application is seeing the latency of your presumably lower latency system pool.<br><br>Dean <br><br><br></font><img src="cid:1__=07BB0A91DFE1314A8f9e8a93df938690918c07B@" width="16" height="16" alt="Inactive
hide details for Tejas Rao ---08/01/2016 12:06:15 PM---In my
case GPFS storage is used to store VM images (KVM) and he"><font size="4" color="#424282">Tejas Rao ---08/01/2016 12:06:15 PM---In my case GPFS storage is used to store VM images (KVM) and hence the small IO.</font><font size="4"><br></font><font color="#5F5F5F"><br>From: </font>Tejas Rao <a href="mailto:raot@bnl.gov"><u><font color="#0000FF"><raot@bnl.gov></font></u></a><font color="#5F5F5F"><br>To: </font>gpfsug main discussion list <a href="mailto:gpfsug-discuss@spectrumscale.org"><u><font color="#0000FF"><gpfsug-discuss@spectrumscale.org></font></u></a><font color="#5F5F5F"><br>Date: </font>08/01/2016 12:06 PM<font color="#5F5F5F"><br>Subject: </font>Re: [gpfsug-discuss] HAWC (Highly available write cache)<font color="#5F5F5F"><br>Sent by: </font><a href="mailto:gpfsug-discuss-bounces@spectrumscale.org"><u><font color="#0000FF">gpfsug-discuss-bounces@spectrumscale.org</font></u></a><p><hr width="100%" size="2" align="left" noshade><font size="4"><br><br></font><font size="5"><br>In my case GPFS storage is used to store VM images (KVM) and hence the small IO.<br><br>I always see lots of small 4K writes and the GPFS filesystem block size is 8MB. I thought the reason for the small writes is that the linux kernel requests GPFS to initiate a periodic sync which by default is every 5 seconds and can be controlled by "vm.dirty_writeback_centisecs". <br><br>I thought HAWC would help in such cases and would harden (coalesce) the small writes in the "system" pool and would flush to the "data" pool in larger block size. <br><br>Note - I am not doing direct i/o explicitly. <br><br></font><font size="4"><br></font><font size="5"><br>On 8/1/2016 14:49, Sven Oehme wrote:</font><font size="4"> </font><ul><ul><ul><ul><font size="5">when you say 'synchronous write' what do you mean by that ? <br>if you are talking about using direct i/o (O_DIRECT flag), they don't leverage HAWC data path, its by design.</font><font size="4"><br></font><font size="5"><br>sven</font><font size="4"><br></font><font size="5"><br>On Mon, Aug 1, 2016 at 11:36 AM, Tejas Rao <</font><a href="mailto:raot@bnl.gov" target="_blank"><u><font size="5" color="#0000FF">raot@bnl.gov</font></u></a><font size="5">> wrote:</font><font size="4"> </font><ul><ul><font size="5">I have enabled write cache (HAWC) by running the below commands. The recovery logs are supposedly placed in the replicated system metadata pool (SSDs). I do not have a "system.log" pool as it is only needed if recovery logs are stored on the client nodes.<br><br>mmchfs gpfs01 --write-cache-threshold 64K<br>mmchfs gpfs01 -L 1024M<br>mmchconfig logPingPongSector=no<br><br>I have recycled the daemon on all nodes in the cluster (including the NSD nodes).<br><br>I still see small synchronous writes (4K) from the clients going to the data drives (data pool). I am checking this by looking at "mmdiag --iohist" output. Should they not be going to the system pool?<br><br>Do I need to do something else? How can I confirm that HAWC is working as advertised?<br><br>Thanks.<br><br><br>_______________________________________________<br>gpfsug-discuss mailing list<br>gpfsug-discuss at </font><a href="http://spectrumscale.org/" target="_blank"><u><font size="5" color="#0000FF">spectrumscale.org</font></u></a><u><font size="4" color="#0000FF"><br></font></u><a href="http://gpfsug.org/mailman/listinfo/gpfsug-discuss" target="_blank"><u><font size="5" color="#0000FF">http://gpfsug.org/mailman/listinfo/gpfsug-discuss</font></u></a><font size="4"> </font></ul></ul><font size="5"><br></font><font size="4"><br></font><tt><font size="5"><br>_______________________________________________<br>gpfsug-discuss mailing list<br>gpfsug-discuss at spectrumscale.org</font></tt><u><font size="4" color="#0000FF"><br></font></u><a href="http://gpfsug.org/mailman/listinfo/gpfsug-discuss"><tt><u><font size="5" color="#0000FF">http://gpfsug.org/mailman/listinfo/gpfsug-discuss</font></u></tt></a></ul></ul></ul></ul><tt><font size="4">_______________________________________________<br>gpfsug-discuss mailing list<br>gpfsug-discuss at spectrumscale.org</font></tt><tt><u><font size="4" color="#0000FF"><br></font></u></tt><a href="http://gpfsug.org/mailman/listinfo/gpfsug-discuss"><tt><u><font size="4" color="#0000FF">http://gpfsug.org/mailman/listinfo/gpfsug-discuss</font></u></tt></a><font size="4"><br><br><br><br><br></font><br><tt><font size="4">_______________________________________________<br>gpfsug-discuss mailing list<br>gpfsug-discuss at spectrumscale.org<br></font></tt><a href="http://gpfsug.org/mailman/listinfo/gpfsug-discuss"><tt><u><font size="4" color="#0000FF">http://gpfsug.org/mailman/listinfo/gpfsug-discuss</font></u></tt></a><tt><font size="4"><br></font></tt></ul></ul><tt>_______________________________________________<br>gpfsug-discuss mailing list<br>gpfsug-discuss at spectrumscale.org<br></tt><tt><a href="http://gpfsug.org/mailman/listinfo/gpfsug-discuss">http://gpfsug.org/mailman/listinfo/gpfsug-discuss</a></tt><tt><br></tt><br><br><BR>
</body></html>