Hi list, I want to monitor the read and write ops/bandwidth for a couple of pools and I''m not quite sure how to proceed. I''m using rrdtool so I either want an accumulated counter or a gauge. According to the ZFS admin guide, running zpool iostat without any parameters should show the activity since boot. On my system (OSOL snv_133) it''s only showing ops in the single digits for a system with a months uptime and many GB of transfers. So, is there a way to get this output correctly, or is there a better way to do this? Thanks
Matt Harrison <iwasinnamuknow at genestate.com> wrote:>Hi list, > >I want to monitor the read and write ops/bandwidth for a couple of >pools >and I''m not quite sure how to proceed. I''m using rrdtool so I either >want an accumulated counter or a gauge. > >According to the ZFS admin guide, running zpool iostat without any >parameters should show the activity since boot. On my system (OSOLAverage activity since boot...>snv_133) it''s only showing ops in the single digits for a system with a > >months uptime and many GB of transfers. > >So, is there a way to get this output correctly, or is there a better >way to do this? > >Thanks >_______________________________________________ >zfs-discuss mailing list >zfs-discuss at opensolaris.org >http://mail.opensolaris.org/mailman/listinfo/zfs-discuss-- Sent from my Android phone with K-9 Mail. Please excuse my brevity.
On 28/06/2011 16:44, Tomas ?gren wrote:> > > Matt Harrison<iwasinnamuknow at genestate.com> wrote: > >> Hi list, >> >> I want to monitor the read and write ops/bandwidth for a couple of >> pools >> and I''m not quite sure how to proceed. I''m using rrdtool so I either >> want an accumulated counter or a gauge. >> >> According to the ZFS admin guide, running zpool iostat without any >> parameters should show the activity since boot. On my system (OSOL > > Average activity since boot...Ahh ok, perhaps the guide should be updated to reflect this.> >> snv_133) it''s only showing ops in the single digits for a system with a >> >> months uptime and many GB of transfers. >> >> So, is there a way to get this output correctly, or is there a better >> way to do this? >> >> ThanksThank you
2011-06-28 19:51, Matt Harrison ?????:> On 28/06/2011 16:44, Tomas ?gren wrote: >> >> >> Matt Harrison<iwasinnamuknow at genestate.com> wrote: >> >>> Hi list, >>> >>> I want to monitor the read and write ops/bandwidth for a couple of >>> pools >>> and I''m not quite sure how to proceed. I''m using rrdtool so I either >>> want an accumulated counter or a gauge. >>> >>> According to the ZFS admin guide, running zpool iostat without anyAlas, I do not know a direct answer, but possibly some kernel stats (like those used by iostat/vmstat to produce disk io statistics) should be informative. I am not sure if there are kstats for pools though. Note that regularly running "zpool iostat" may tax your system resources to spawn a process, as well as lock up the system if there are any problems with the pools and the kernel is waiting on something indefinitely (i.e. my loooong pool imports for which the kernel needed some sanity checks not unlike full zdb walks) - during these hours and days any zfs/zpool/bootadm/df commands were unuseable, and if you run them regularly a-la crontab, you can end up with thousands of hanging processes and deplete any RAM you can have. Hint: make a script to use lock-files to avoid spawning more than one "zpool iostat" request from rrdtools. HTH, //Jim -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20110628/bf064c4c/attachment.html>
FYI, I found the following article helpful in setting up monitoring for my home setup: http://www.markround.com/archives/48-Linux-iostat-monitoring-with-Cacti.html The article above is written in the context of setting up monitoring via Cacti, but that''s simply using rrdtool on the backend and its the same general idea. Basically: # Setup a cronjob to runs every 5 minutes that runs "iostat -dx 30 2" to gather I/O stats for 30 seconds. You could increase the poll-time if you wanted a more accurate snapshot, but getting statistics for a given 30-second block at a time seemed granular enough for my needs. # Parse the results and feed into rrdtool. Since I''m doing remote monitoring, I used the perl script from the above article to create an net-snmp extension that would return the parsed results, so that Cacti can query the stats over SNMP. This gives the I/O stats on a per drive basis (rather than overall pool stats), but you can adapt the idea to whatever you need. -Tony On Tue, Jun 28, 2011 at 10:08 AM, Matt Harrison < iwasinnamuknow at genestate.com> wrote:> Hi list, > > I want to monitor the read and write ops/bandwidth for a couple of pools > and I''m not quite sure how to proceed. I''m using rrdtool so I either want an > accumulated counter or a gauge. > > According to the ZFS admin guide, running zpool iostat without any > parameters should show the activity since boot. On my system (OSOL snv_133) > it''s only showing ops in the single digits for a system with a months uptime > and many GB of transfers. > > So, is there a way to get this output correctly, or is there a better way > to do this? > > Thanks > ______________________________**_________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/**mailman/listinfo/zfs-discuss<http://mail.opensolaris.org/mailman/listinfo/zfs-discuss> >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20110629/9c53e382/attachment.html>