| bio | website | |
|---|---|---|
| location | ||
| age | ||
| visits | member for | 1 year, 2 months |
| seen | May 10 at 13:28 | |
| stats | profile views | 2 |
|
Mar 27 |
revised |
PostgreSQL slow commit performance added 360 characters in body |
|
Mar 27 |
revised |
PostgreSQL slow commit performance added 2049 characters in body |
|
Mar 27 |
comment |
PostgreSQL slow commit performance I have amended my original question with additional information about mount options and partition alignment. |
|
Mar 27 |
revised |
PostgreSQL slow commit performance added 1624 characters in body |
|
Mar 27 |
comment |
PostgreSQL slow commit performance I ran bonnie++ which shows good performance (as good as you'd expect from regular sata disks). Also, the partitions are aligned. When I ran pg_test_fsync the first time it was on a VM. Then I ran it on the actual hardware, after shutting down every other process (including VMs). The performance was slightly better, around 40 ops/sec, which is still deplorable. I'm going to run some more tests, on separate partitions if I got the time today. Thanks for all the suggestions. |
|
Mar 25 |
comment |
PostgreSQL slow commit performance Just checked, the firmware on both disks is at the latest version (CC4H). |
|
Mar 25 |
comment |
PostgreSQL slow commit performance I have amended my original question with information about the disk types. In the meantime I tried setting noatime and killed all other processes (when possible) on the machine. This increases the performance two-fold, from extremely bad, to just really bad. |
|
Mar 25 |
revised |
PostgreSQL slow commit performance added 150 characters in body |
|
Mar 25 |
awarded | Commentator |
|
Mar 25 |
comment |
PostgreSQL slow commit performance The short smart test on both disks revealed no problems, I started long tests on both, and will check the results asap. dmesg and the syslog showed no errors. I checked the partition aligned friday, they are aligned on 4k boundaries, also, the disks use GPT, so fdisk -l doesn't show useful output. |
|
Mar 22 |
revised |
PostgreSQL slow commit performance added 44 characters in body |
|
Mar 22 |
comment |
PostgreSQL slow commit performance I have updated the question with more details about the filesystem and raid config. I understand that this machine is never going to give very good performance in it's current configuration. But the current performance is really bad. |
|
Mar 22 |
revised |
PostgreSQL slow commit performance Formatting |
|
Mar 22 |
comment |
PostgreSQL slow commit performance I tried running pg_test_fsync on my own SSD, and I get comparable performance figures. I know that I could disable sync commits, but the question remains, what is the cause of this problem? |
|
Mar 22 |
comment |
PostgreSQL slow commit performance Yes, but what is causing the bad fsync performance? |
|
Mar 22 |
revised |
PostgreSQL slow commit performance added 75 characters in body |
|
Mar 22 |
asked | PostgreSQL slow commit performance |
|
Apr 6 |
awarded | Teacher |
|
Apr 5 |
answered | FreeBSD network performance fluctuating |
|
Apr 5 |
revised |
FreeBSD network performance fluctuating added 424 characters in body |