Postgres, fsync and RAID controller with 100M of internal cache & dedicated battery

["Dmitry Koterov" <dmitry@xxxxxxxxxx>]

And here are results of built-in Postgres test script:

Simple write timing:
        write                    0.006355

Compare fsync times on write() and non-write() descriptor:
(If the times are similar, fsync() can sync data written
 on a different descriptor.)
        write, fsync, close      0.233793
        write, close, fsync      0.227444

Compare one o_sync write to two:
        one 16k o_sync write     0.297093
        two 8k o_sync writes     
0.402803

Compare file sync methods with one 8k write:

        (o_dsync unavailable)
        write, fdatasync         0.228725
        write, fsync,            0.223302

Compare file sync methods with 2 8k writes:
        (o_dsync unavailable)
        open o_sync, write       0.414954
        write, fdatasync         0.335280
        write, fsync,            0.327195

(Also, I tried to manually specify open_sync method in 
postgresql.conf, but after that Postgres database had completely crashed. :-)

On 8/22/07, Dmitry Koterov <dmitry@xxxxxxxxxx > wrote:

All settings seems to be fine. Mode is writeback.

We temporarily (for tests only on test machine!!!) put pg_xlog into RAM drive (to completely exclude xlog fsync from the statistics), but slowdown during the checkpoint and 5-10 second fsync during the checkpoint are alive yet.

Here are some statistical data from the controller. Other report data is attached to the mail.

ACCELERATOR STATUS:
   Logical Drive Disable Map: 0x00000000
   Read Cache Size:           24 MBytes
   Posted Write Size:         72 MBytes
   Disable Flag:              0x00
   Status:                    0x00000001
   Disable Code:              0x0000
   Total Memory Size:         128 MBytes
   Battery Count:             1
   Battery Status:            0x0001
   Parity Read Errors:        0000
   Parity Write Errors:       0000
   Error Log:                 N/A
   Failed Batteries:          0x0000
   Board Present:             Yes
   Accelerator Failure Map:   0x00000000
   Max Error Log Entries:     12
   NVRAM Load Status:         0x00
   Memory Size Shift Factor:  0x0a
   Non Battery Backed Memory: 0 MBytes
   Memory State:              0x00

On 8/22/07, Scott Marlowe <scott.marlowe@xxxxxxxxx> wrote:

On 8/22/07, Dmitry Koterov <dmitry@xxxxxxxxxx> wrote:
> Hello.
> You see, 50M block was fsynced for 0.25 s.
>
> The question is: how to solve this problem and make fsync run with no delay.
> Seems to me that controller's internal write cache is not used (strange,
> because all configuration options are fine), but how to check it? Or, maybe,
> there is another side-effect?

I would suggest that either the controller is NOT configured fine, OR
there's some bug in how the OS is interacting with it.

What options are there for this RAID controller, and what are they set
to?  Specifically, the writeback / writethru type options for the
cache, and it might be if it doesn't preoprly detect a battery backup
module it refuses to go into writeback mode.