Date: Mon, 12 Jun 2017 14:13:30 +0800 From: Jov <amutu@amutu.com> To: "Caza, Aaron" <Aaron.Caza@ca.weatherford.com> Cc: "freebsd-hackers@freebsd.org" <freebsd-hackers@freebsd.org>, freebsd-fs <freebsd-fs@freebsd.org> Subject: Re: [EXTERNAL] Re: FreeBSD10 Stable + ZFS + PostgreSQL + SSD performance drop < 24 hours Message-ID: <CADyrUxPH3LPOePhvz3gFyvWbAHrx%2BwU04pcrpBhKXDvF4=p0Pw@mail.gmail.com> In-Reply-To: <4410c303314a4d11832a8d248e0b53e1@DM2PR58MB013.032d.mgd.msft.net> References: <a8523e8099404bd699525f8ff7763819@DM2PR58MB013.032d.mgd.msft.net> <CADyrUxNeRW_fKJzoc9zoZi3J-6L%2BtXnJuW3q_4cremUy7sP7fA@mail.gmail.com> <4410c303314a4d11832a8d248e0b53e1@DM2PR58MB013.032d.mgd.msft.net>
next in thread | previous in thread | raw e-mail | index | archive | help
>From the output of explain analyze of PG, the problem can be excluded from the database.I am not a fs expert, I CCed freebsd-fs@freebsd.org.It may be helpful if you provide more info such as sysctl -a | grep zfs after degradation. 2017-06-12 12:50 GMT+08:00 Caza, Aaron <Aaron.Caza@ca.weatherford.com>: > Thanks, Jov, for you suggestions. Per your e-mail I added =E2=80=9Cexpla= in > analyze=E2=80=9D to the script: > > > > #!/bin/sh > psql --username=3Dtest --password=3Dsupersecret -h /db -d test << EOL > \timing on > explain analyze select count(*) from test; > \q > EOL > > Sample run of above script before degradation: > > Timing is on. > > QUERY > PLAN > > ------------------------------------------------------------ > ------------------------------------------------------------------------ > > Aggregate (cost=3D3350822.35..3350822.36 rows=3D1 width=3D0) (actual > time=3D60234.556..60234.556 rows=3D1 loops=3D1) > > -> Seq Scan on test (cost=3D0.00..3296901.08 rows=3D21568508 width= =3D0) > (actual time=3D1.126..57021.470 rows=3D21568508 loops=3D1) > > Planning time: 4.968 ms > > Execution time: 60234.649 ms > > (4 rows) > > > > Time: 60248.503 ms > > test$ uptime > > 10:33PM up 7 mins, 3 users, load averages: 1.68, 1.79, 0.94 > > > > > > Sample run of above script after degradation (~11.33 hours uptime): > > Timing is on. > > QUERY > PLAN > > ------------------------------------------------------------ > ------------------------------------------------------------------------- > > Aggregate (cost=3D3350822.35..3350822.36 rows=3D1 width=3D0) (actual > time=3D485669.361..485669.361 rows=3D1 loops=3D1) > > -> Seq Scan on test (cost=3D0.00..3296901.08 rows=3D21568508 width= =3D0) > (actual time=3D0.008..483241.253 rows=3D21568508 loops=3D1) > > Planning time: 0.529 ms > > Execution time: 485669.411 ms > > (4 rows) > > > > Time: 485670.432 ms > > test$ uptime > > 9:59PM up 11:21, 2 users, load averages: 1.11, 2.13, 2.14 > > > > > > Regarding dd=E2=80=99ing the pgdata directory, that didn=E2=80=99t work f= or me as Postgres > splits the database up into multiple 2GB files =E2=80=93 dd=E2=80=99ing o= f a 2GB file on a > system with 8GB ram doesn=E2=80=99t seem representative. I opted to crea= te a 16GB > file (dd if=3D/dev/random of=3D/testdb/test bs=3D1m count=3D16000) on th= e > pertinent ZFS file system then performed dd operation on that: > > > > Sample of run after degradation (~11.66 hours uptime): > > 16000+0 records in > > 16000+0 records out > > 16777216000 bytes transferred in 274.841792 secs (61043176 bytes/sec) > > test$ uptime > > 10:25PM up 11:46, 2 users, load averages: 1.00, 1.28, 1.59 > > > > > > After rebooting, we can see **MUCH** before performance: > > test$ dd if=3D/testdb/test of=3D/dev/null bs=3D1m > > 16000+0 records in > > 16000+0 records out > > 16777216000 bytes transferred in 19.456043 secs (862313883 bytes/sec) > > test$ dd if=3D/testdb/test of=3D/dev/null bs=3D1m > > 16000+0 records in > > 16000+0 records out > > 16777216000 bytes transferred in 19.375321 secs (865906473 bytes/sec) > > test$ dd if=3D/testdb/test of=3D/dev/null bs=3D1m > > 16000+0 records in > > 16000+0 records out > > 16777216000 bytes transferred in 19.173458 secs (875022968 bytes/sec) > > test$ uptime > > 10:30PM up 4 mins, 3 users, load averages: 3.52, 1.62, 0.69 > > > > These tests were conducted with the previously mentioned Samsung 850 Pro > 256GB SSDs (Intel Xeon E31240 with 8GB ram). There=E2=80=99s essentially= nothing > else running on this system (99.5-100% idle) and no other disk activity. > > > > Regards, > > A > > > > *From:* Jov [mailto:amutu@amutu.com] > *Sent:* Sunday, June 11, 2017 5:50 PM > *To:* Caza, Aaron > *Cc:* freebsd-hackers@freebsd.org; Allan Jude > > *Subject:* [EXTERNAL] Re: FreeBSD10 Stable + ZFS + PostgreSQL + SSD > performance drop < 24 hours > > > > To exclude the fs problem=EF=BC=8CI will do a dd test on the pgdata data = set > after the performance drop,if the read and/or write utility can reach 100= % > or performance expected then I will say the problem is not fs or os. > > > > For pg,what's your output of explain analyze before and after performance > drop? > > > > 2017=E5=B9=B46=E6=9C=8812=E6=97=A5 12:51 AM=EF=BC=8C"Caza, Aaron" <Aaron.= Caza@ca.weatherford.com>=E5=86=99=E9=81=93=EF=BC=9A > > Thanks Allan for the suggestions. I tried gstat -d but deletes (d/s) > doesn't seem to be it as it stays at 0 despite vfs.zfs.trim.enabled=3D1. > > This is most likely due to the "layering" I use as, for historical > reasons, I have GEOM ELI set up to essentially emulate 4k sectors > regardless of the underlying media. I do my own alignment and partition > sizing as well as have the ZFS record size set to 8k for Postgres. > > In gstat, the SSDs %busy is 90-100% on startup after reboot. Once the > performance degradation hits (<24 hours later), I'm seeing %busy at ~10%. > > #!/bin/sh > psql --username=3Dtest --password=3Dsupersecret -h /db -d test << EOL > \timing on > select count(*) from test; > \q > EOL > > Sample run of above script after reboot (before degradation hits) (Samsun= g > 850 Pros in ZFS mirror): > Timing is on. > count > ---------- > 21568508 > (1 row) > > Time: 57029.262 ms > > Sample run of above script after degradation (Samsung 850 Pros in ZFS > mirror): > Timing is on. > count > ---------- > 21568508 > (1 row) > > Time: 583595.239 ms > (Uptime ~1 day in this particular case.) > > > Any other suggestions? > > Regards, > A > > -----Original Message----- > From: owner-freebsd-hackers@freebsd.org [mailto:owner-freebsd-hackers@ > freebsd.org] On Behalf Of Allan Jude > Sent: Saturday, June 10, 2017 9:40 PM > To: freebsd-hackers@freebsd.org > Subject: [EXTERNAL] Re: FreeBSD10 Stable + ZFS + PostgreSQL + SSD > performance drop < 24 hours > > On 06/10/2017 12:36, Slawa Olhovchenkov wrote: > > On Sat, Jun 10, 2017 at 04:25:59PM +0000, Caza, Aaron wrote: > > > >> Gents, > >> > >> I'm experiencing an issue where iterating over a PostgreSQL table of > ~21.5 million rows (select count(*)) goes from ~35 seconds to ~635 second= s > on Intel 540 SSDs. This is using a FreeBSD 10 amd64 stable kernel back > from Jan 2017. SSDs are basically 2 drives in a ZFS mirrored zpool. I'm > using PostgreSQL 9.5.7. > >> > >> I've tried: > >> > >> * Using the FreeBSD10 amd64 stable kernel snapshot of May 25, > 2017. > >> > >> * Tested on half a dozen machines with different models of SSDs: > >> > >> o Intel 510s (120GB) in ZFS mirrored pair > >> > >> o Intel 520s (120GB) in ZFS mirrored pair > >> > >> o Intel 540s (120GB) in ZFS mirrored pair > >> > >> o Samsung 850 Pros (256GB) in ZFS mirrored pair > >> > >> * Using bonnie++ to remove Postgres from the equation and > performance does indeed drop. > >> > >> * Rebooting server and immediately re-running test and > performance is back to original. > >> > >> * Tried using Karl Denninger's patch from PR187594 (which took > some work to find a kernel that the FreeBSD10 patch would both apply and > compile cleanly against). > >> > >> * Tried disabling ZFS lz4 compression. > >> > >> * Ran the same test on a FreeBSD9.0 amd64 system using PostgreSQ= L > 9.1.3 with 2 Intel 520s in ZFS mirrored pair. System had 165 days uptime > and test took ~80 seconds after which I rebooted and re-ran test and was > still at ~80 seconds (older processor and memory in this system). > >> > >> I realize that there's a whole lot of info I'm not including (dmesg, > zfs-stats -a, gstat, et cetera): I'm hoping some enlightened individual > will be able to point me to a solution with only the above to go on. > > > > Just a random guess: can you try r307264 (I am mean regression in > > r307266)? > > _______________________________________________ > > freebsd-hackers@freebsd.org mailing list > > https://lists.freebsd.org/mailman/listinfo/freebsd-hackers > > To unsubscribe, send any mail to "freebsd-hackers-unsubscribe@ > freebsd.org" > > > > This sounds a bit like an issue I investigated for a customer a few month= s > ago. > > Look at gstat -d (includes DELETE operations like TRIM) > > If you see a lot of that happening, but try: vfs.zfs.trim.enabled=3D0 in > /boot/loader.conf and see if your issues go away. > > the FreeBSD TRIM code for ZFS basicallys waits until the sector has been > free for a while (to avoid doing a TRIM on a block we'll immediately > reuse), so your benchmark will run file for a little while, then suddenly > the TRIM will kick in. > > For postgres, fio, bonnie++ etc, make sure the ZFS dataset you are storin= g > the data on / benchmarking has a recordsize that matches the workload. > > If you are doing a write-only benchmark, and you see lots of reads in > gstat, you know you are having to do read/modify/write's, and that is why > your performance is so bad. > > > -- > Allan Jude > _______________________________________________ > freebsd-hackers@freebsd.org mailing list https://lists.freebsd.org/ > mailman/listinfo/freebsd-hackers > To unsubscribe, send any mail to "freebsd-hackers-unsubscribe@freebsd.org= " > > This message may contain confidential and privileged information. If it > has been sent to you in error, please reply to advise the sender of the > error and then immediately delete it. If you are not the intended > recipient, do not read, copy, disclose or otherwise use this message. The > sender disclaims any liability for such unauthorized use. PLEASE NOTE tha= t > all incoming e-mails sent to Weatherford e-mail accounts will be archived > and may be scanned by us and/or by external service providers to detect a= nd > prevent threats to our systems, investigate illegal or inappropriate > behavior, and/or eliminate unsolicited promotional e-mails (spam). This > process could result in deletion of a legitimate e-mail before it is read > by its intended recipient at our organization. Moreover, based on the > scanning results, the full text of e-mails and attachments may be made > available to Weatherford security and other personnel for review and > appropriate action. If you have any concerns about this process, > please contact us at dataprivacy@weatherford.com. > _______________________________________________ > freebsd-hackers@freebsd.org mailing list > https://lists.freebsd.org/mailman/listinfo/freebsd-hackers > To unsubscribe, send any mail to "freebsd-hackers-unsubscribe@freebsd.org= " >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CADyrUxPH3LPOePhvz3gFyvWbAHrx%2BwU04pcrpBhKXDvF4=p0Pw>