[RFC] md raid resync counter

[RFC] md raid resync counter

[Jack Wang]

Hi Neil and all,

Is there anyway to know:

1. How often does a raid start to resync?
2. How many bytes did it resync?

If not, for the first one, I think we cound add field into mddev like
resync_counter, increase when resync done, and annother field into
md_rdev to track how many bytes synced to the outdated disk?

Regards,
Jack

Re: [RFC] md raid resync counter

[NeilBrown]

[-- Attachment #1: Type: text/plain, Size: 548 bytes --]

On Fri, 01 Nov 2013 13:59:40 +0100 Jack Wang <xjtuwjp@gmail.com> wrote:

> Hi Neil and all,
> 
> Is there anyway to know:
> 
> 1. How often does a raid start to resync?

Look at your logs?

> 2. How many bytes did it resync?

Look at "mismatch_cnt" in sysfs (and multiply by 512).

NeilBrown


> 
> If not, for the first one, I think we cound add field into mddev like
> resync_counter, increase when resync done, and annother field into
> md_rdev to track how many bytes synced to the outdated disk?
> 
> Regards,
> Jack


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: [RFC] md raid resync counter

[Jack Wang]

On 11/02/2013 12:48 AM, NeilBrown wrote:
> On Fri, 01 Nov 2013 13:59:40 +0100 Jack Wang <xjtuwjp@gmail.com> wrote:
> 
>> Hi Neil and all,
>>
>> Is there anyway to know:
>>
>> 1. How often does a raid start to resync?
> 
> Look at your logs?
> 
>> 2. How many bytes did it resync?
> 
> Look at "mismatch_cnt" in sysfs (and multiply by 512).
> 
> NeilBrown
> 
>
Thanks Neil,

You're always very helpful :)

About the second one. As I checked in my system kernel 3.4.51

when raid is resync:
cat /proc/mdstat
Personalities : [raid1]
md1 : active raid1 dm-1[3] dm-2[2]
      23762944 blocks super 1.2 [2/1] [_U]
      [==>..................]  recovery = 10.7% (2562240/23762944)
finish=7.5min speed=46565K/sec
# cat /sys/block/md1/md/mismatch_cnt
0

cat /sys/block/md1/md/degraded
1
 cat /sys/block/md1/md/mismatch_cnt
0
# cat /sys/block/md1/md/sync_completed
8911488 / 47525888

mismatch_cnt is 0, sync_completed look what I want?

Jack

Re: [RFC] md raid resync counter

[NeilBrown]

[-- Attachment #1: Type: text/plain, Size: 1805 bytes --]

On Mon, 04 Nov 2013 10:26:42 +0100 Jack Wang <xjtuwjp@gmail.com> wrote:

> On 11/02/2013 12:48 AM, NeilBrown wrote:
> > On Fri, 01 Nov 2013 13:59:40 +0100 Jack Wang <xjtuwjp@gmail.com> wrote:
> > 
> >> Hi Neil and all,
> >>
> >> Is there anyway to know:
> >>
> >> 1. How often does a raid start to resync?
> > 
> > Look at your logs?
> > 
> >> 2. How many bytes did it resync?
> > 
> > Look at "mismatch_cnt" in sysfs (and multiply by 512).
> > 
> > NeilBrown
> > 
> >
> Thanks Neil,
> 
> You're always very helpful :)
> 
> About the second one. As I checked in my system kernel 3.4.51
> 
> when raid is resync:
> cat /proc/mdstat
> Personalities : [raid1]
> md1 : active raid1 dm-1[3] dm-2[2]
>       23762944 blocks super 1.2 [2/1] [_U]
>       [==>..................]  recovery = 10.7% (2562240/23762944)
> finish=7.5min speed=46565K/sec

This array is doing "recovery", not "resync".

"Resync" is when the devices should all contains the same data but maybe
don't due to some  error.  So md/raid1 checks and fixes the errors.
"recovery" is when a spare has been added and data is being copied  onto it.


> # cat /sys/block/md1/md/mismatch_cnt
> 0
> 
> cat /sys/block/md1/md/degraded
> 1
>  cat /sys/block/md1/md/mismatch_cnt
> 0
> # cat /sys/block/md1/md/sync_completed
> 8911488 / 47525888
> 
> mismatch_cnt is 0, sync_completed look what I want?

"sync_completed" is a number which is somewhat smaller than the number of
sectors that have been be processed by the current
sync/repair/recover/whatever.

Slightly smaller because it is only updated occasionally.

I don't really know what you want because you didn't say what you would do
with the information, but it is possible that this is what you want.


> 
> Jack


NeilBrown

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: [RFC] md raid resync counter

[Jack Wang]

On 11/05/2013 06:12 AM, NeilBrown wrote:
> On Mon, 04 Nov 2013 10:26:42 +0100 Jack Wang <xjtuwjp@gmail.com> wrote:
> 
>> On 11/02/2013 12:48 AM, NeilBrown wrote:
>>> On Fri, 01 Nov 2013 13:59:40 +0100 Jack Wang <xjtuwjp@gmail.com> wrote:
>>>
>>>> Hi Neil and all,
>>>>
>>>> Is there anyway to know:
>>>>
>>>> 1. How often does a raid start to resync?
>>>
>>> Look at your logs?
>>>
>>>> 2. How many bytes did it resync?
>>>
>>> Look at "mismatch_cnt" in sysfs (and multiply by 512).
>>>
>>> NeilBrown
>>>
>>>
>> Thanks Neil,
>>
>> You're always very helpful :)
>>
>> About the second one. As I checked in my system kernel 3.4.51
>>
>> when raid is resync:
>> cat /proc/mdstat
>> Personalities : [raid1]
>> md1 : active raid1 dm-1[3] dm-2[2]
>>       23762944 blocks super 1.2 [2/1] [_U]
>>       [==>..................]  recovery = 10.7% (2562240/23762944)
>> finish=7.5min speed=46565K/sec
> 
> This array is doing "recovery", not "resync".
> 
> "Resync" is when the devices should all contains the same data but maybe
> don't due to some  error.  So md/raid1 checks and fixes the errors.
> "recovery" is when a spare has been added and data is being copied  onto it.
> 
> 
>> # cat /sys/block/md1/md/mismatch_cnt
>> 0
>>
>> cat /sys/block/md1/md/degraded
>> 1
>>  cat /sys/block/md1/md/mismatch_cnt
>> 0
>> # cat /sys/block/md1/md/sync_completed
>> 8911488 / 47525888
>>
>> mismatch_cnt is 0, sync_completed look what I want?
> 
> "sync_completed" is a number which is somewhat smaller than the number of
> sectors that have been be processed by the current
> sync/repair/recover/whatever.
> 
> Slightly smaller because it is only updated occasionally.
> 
> I don't really know what you want because you didn't say what you would do
> with the information, but it is possible that this is what you want.

Thanks Neil for clarification.
I think we need track both recovery and resync status, mismatch_cnt and
sync_completed are what we want.

Regards,
Jack
> 
> 
>>
>> Jack
> 
> 
> NeilBrown
>