Re: [PATCH 1/5] btrfs-progs: scan /proc/partitions not all of /dev with "-d"

From: Anand Jain <Anand.Jain@oracle.com>
To: linux-btrfs@vger.kernel.org
Cc: dsterba@suse.cz, guihc.fnst@cn.fujitsu.com, sandeen@redhat.com
Subject: Re: [PATCH 1/5] btrfs-progs: scan /proc/partitions not all of /dev with "-d"
Date: Thu, 23 Oct 2014 21:12:48 +0800	[thread overview]
Message-ID: <5448FED0.7030409@oracle.com> (raw)
In-Reply-To: <5432617B.8050301@oracle.com>

  my stap func profiling script was wrong, I got the number of
  times scan_lblkid func called per thread wrong, now its
  been corrected as below. yet calling the system-wide device
  scan more than once per thread does not make any sense. There
  are quite a number of threads like that as below. The worst is
  mkfs.btrfs which calls n number of times, where n is number of
  disk being mkfs-ed.

   btrfs-find-root            1
   btrfs rescue super-recover 2
   btrfs-debug-tree           1
   btrfs-image -r             2
   btrfs check                2
   btrfs restore              2
   calc-size                  NC
   btrfs-corrupt-block        NC
   btrfs-image                NC
   btrfs-map-logical          1
   btrfs-select-super         NC
   btrfstune                  2
   btrfs-zero-log             NC
   tester                     NC
   quick-test.c               NC
   btrfs-convert              0
   mkfs                       #number of devices to be mkfs
   btrfs label set unmounted  2
   btrfs get label unmounted  2

On 10/06/14 17:31, Anand Jain wrote:
>
>
>   I am running some tests with larger disks pool (48 disks).
>   With that the performance of the various scan methods are as below..
>
> ----
> scanning BTRFS_SCAN_LBLKID
> real    0m0.330s
> user    0m0.005s
> sys     0m0.026s
>
> scanning BTRFS_SCAN_DEV
> real    0m0.010s
> user    0m0.002s
> sys     0m0.005s
>
> scanning BTRFS_SCAN_PROC
> real    0m0.010s
> user    0m0.002s
> sys     0m0.005s
> -----
>
>   This is the time taken to scan 48disks one time by various methods
>   we have/had - but our progs do this scan 30times for btrfs fi show.
>   yep 30times as show below.. I am working to fix it.
>
> -------
>                                  Function: time(us)    count  avg(us)
> ::
>                           get_device_info:     1034       27       38
>                      pretty_size_snprintf:     1218      124        9
>                     btrfs_scan_one_device:     1790      186        9
>                      btrfs_read_dev_super:     1956      116       16
>                                  cmd_show:    15418      335       46
>                         btrfs_scan_lblkid:   148477       30     4949
> -------
>
>   IMO we should still stick to LBLKID scan.
>   Just a thought - any idea if its better to provide a compile time
>   fallback scan switch, just in case if something fails with lblkid. ?
>
>
> Thanks, Anand
>
>
> On 09/24/14 01:00, David Sterba wrote:
>> Hi,
>>
>> all 5 patches will be in the next integration. I haven't tested them
>> yet, seems it's a bit more important to make a more stable devel base
>> for more updates you might want to send.
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html