dm-cache issue

dm-cache issue

[Alexander Pashaliyski]

[-- Attachment #1.1: Type: text/plain, Size: 663 bytes --]

Hi guys,


I am in a process of evaluating dm-cache for our backup system.

Currently I have an issue when restart the backup server. The server is
booting for hours, because of IO load. It seems is triggered a flush from
SSD disk (that is used for a cache device) to the raid controllers (they
are with slow SATA disks).
I have 10 cached logical volumes in *writethrough mode*, each with 2T of
data over 2 raid controllers. I use a single SSD disk for the cache.
The backup system is with lvm2-2.02.164-1 & kernel 4.4.30.


Do you have any ideas why such flush is triggered? In writethrough cache
mode we shouldn't have dirty blocks in the cache.


Thanks,
Alex

[-- Attachment #1.2: Type: text/html, Size: 772 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]

Re: dm-cache issue

[Zdenek Kabelac]

Dne 14.11.2016 v 16:02 Alexander Pashaliyski napsal(a):
> Hi guys,
>
>
> I am in a process of evaluating dm-cache for our backup system.
>
> Currently I have an issue when restart the backup server. The server is
> booting for hours, because of IO load. It seems is triggered a flush from SSD
> disk (that is used for a cache device) to the raid controllers (they are with
> slow SATA disks).
> I have 10 cached logical volumes in *writethrough mode*, each with 2T of data
> over 2 raid controllers. I use a single SSD disk for the cache.
> The backup system is with lvm2-2.02.164-1 & kernel 4.4.30.
>
>
> Do you have any ideas why such flush is triggered? In writethrough cache mode
> we shouldn't have dirty blocks in the cache.
>

Hi

Have you ensured there was proper shutdown ?
Cache needs to be properly deactivated - if it's just turned off,
all metadata are marked dirty.


Regards

Zdenek

Re: dm-cache issue

[Alexander Pashaliyski]

On 11/14/2016 05:34 PM, Zdenek Kabelac wrote:
> Dne 14.11.2016 v 16:02 Alexander Pashaliyski napsal(a):
>> Hi guys,
>>
>>
>> I am in a process of evaluating dm-cache for our backup system.
>>
>> Currently I have an issue when restart the backup server. The server is
>> booting for hours, because of IO load. It seems is triggered a flush 
>> from SSD
>> disk (that is used for a cache device) to the raid controllers (they 
>> are with
>> slow SATA disks).
>> I have 10 cached logical volumes in *writethrough mode*, each with 2T 
>> of data
>> over 2 raid controllers. I use a single SSD disk for the cache.
>> The backup system is with lvm2-2.02.164-1 & kernel 4.4.30.
>>
>>
>> Do you have any ideas why such flush is triggered? In writethrough 
>> cache mode
>> we shouldn't have dirty blocks in the cache.
>>
>
> Hi
>
> Have you ensured there was proper shutdown ?
> Cache needs to be properly deactivated - if it's just turned off,
> all metadata are marked dirty.
>
>
> Regards
>
> Zdenek
>
Hi Zdenek,

Thank you for your answer.
Yes, I am sure. I have modified the lvm2 init script to de-activate all 
logical volumes and then volume groups, but the issue still persist.


Regards,
Alex

Re: dm-cache issue

[Teodor Milkov]

On 14.11.2016 17:34, Zdenek Kabelac wrote:
> Dne 14.11.2016 v 16:02 Alexander Pashaliyski napsal(a):
>> The server is booting for hours, because of IO load. It seems is 
>> triggered a flush from SSD disk (that is used for a cache device) to 
>> the raid controllers (they are with slow SATA disks).
>> I have 10 cached logical volumes in *writethrough mode*, each with 2T 
>> of data over 2 raid controllers. I use a single SSD disk for the cache.
>> The backup system is with lvm2-2.02.164-1 & kernel 4.4.30.
>>
>> Do you have any ideas why such flush is triggered? In writethrough 
>> cache mode
>> we shouldn't have dirty blocks in the cache.
>>
>
> Have you ensured there was proper shutdown ?
> Cache needs to be properly deactivated - if it's just turned off,
> all metadata are marked dirty.
>
> Zdenek

Hi,

I'm seeing the same behavior described by Alexander. Even if we assume 
something is wrong with my shutdown scripts, still how could dm-cache 
ever be dirty in writethrough mode? What about the case where server 
crashes for whatever reason (kernel bug, power outage, operator error 
etc.)? Waiting several hours, or for sufficiently large cache even days 
for the system to come back up is not practical.

I found this 2013 conversation, where Heinz Mauelshagen <heinzm redhat 
com> states that "in writethrough mode the cache will always be coherent 
after a crash": 
https://www.redhat.com/archives/dm-devel/2013-July/msg00117.html

I'm thinking for a way to --uncache and recreate cache devices on every 
boot, which should be safe in writethrough mode and takes reasonable, 
and more importantly – constant amount of time.

Best regards,
Teodor Milkov

--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel

Re: dm-cache issue

[John Stoffel]

Alexander> I am in a process of evaluating dm-cache for our backup system.

Why are you bothering to cache your backup destination?  What
filesystems are you using for your destinations?

Alexander> Currently I have an issue when restart the backup
Alexander> server. The server is booting for hours, because of IO
Alexander> load. It seems is triggered a flush from SSD disk (that is
Alexander> used for a cache device) to the raid controllers (they are
Alexander> with slow SATA disks).  I have 10 cached logical volumes in
Alexander> writethrough mode, each with 2T of data over 2 raid
Alexander> controllers. I use a single SSD disk for the cache.  The
Alexander> backup system is with lvm2-2.02.164-1 & kernel 4.4.30.

How big is the SSD cache?  Do you have any output from the bootup
(syslog, dmesg, etc) you can share.  

Alexander> Do you have any ideas why such flush is triggered? In
Alexander> writethrough cache mode we shouldn't have dirty blocks in
Alexander> the cache.  

I wonder if it makes more sense to use 'lvcache' instead, since you
can add/remove cache from LVs at will.  With bcache, once you setup a
volume with the cache, you're stuck with it.

Also, did you mirror your SSD incase it fails?

In my experience, using RAID6 on a bunch of SATA disks for backup is
fine, you're writing large chunks of data in a few sets of files, so
fragmentation and the Read-Modify-Write cycle is much less painful,
since you tend to stream your writes to the storage, which is
perfectly fine with SATA drives.

Also, bcache skips sequential IO by default, which is what you want
when doing backups... so I'm wondering what your thinking was here for
using this?

John

Re: dm-cache issue

[Zdenek Kabelac]

Dne 15.11.2016 v 13:38 Teodor Milkov napsal(a):
> On 14.11.2016 17:34, Zdenek Kabelac wrote:
>> Dne 14.11.2016 v 16:02 Alexander Pashaliyski napsal(a):
>>> The server is booting for hours, because of IO load. It seems is triggered
>>> a flush from SSD disk (that is used for a cache device) to the raid
>>> controllers (they are with slow SATA disks).
>>> I have 10 cached logical volumes in *writethrough mode*, each with 2T of
>>> data over 2 raid controllers. I use a single SSD disk for the cache.
>>> The backup system is with lvm2-2.02.164-1 & kernel 4.4.30.
>>>
>>> Do you have any ideas why such flush is triggered? In writethrough cache mode
>>> we shouldn't have dirty blocks in the cache.
>>>
>>
>> Have you ensured there was proper shutdown ?
>> Cache needs to be properly deactivated - if it's just turned off,
>> all metadata are marked dirty.
>>
>> Zdenek
>
> Hi,
>
> I'm seeing the same behavior described by Alexander. Even if we assume
> something is wrong with my shutdown scripts, still how could dm-cache ever be
> dirty in writethrough mode? What about the case where server crashes for
> whatever reason (kernel bug, power outage, operator error etc.)? Waiting
> several hours, or for sufficiently large cache even days for the system to
> come back up is not practical.
>
> I found this 2013 conversation, where Heinz Mauelshagen <heinzm redhat com>
> states that "in writethrough mode the cache will always be coherent after a
> crash": https://www.redhat.com/archives/dm-devel/2013-July/msg00117.html
>
> I'm thinking for a way to --uncache and recreate cache devices on every boot,
> which should be safe in writethrough mode and takes reasonable, and more
> importantly – constant amount of time.

My first 'guess' in this reported case is - the disk I/O traffic seen is 
related to the 'reload' of cached chunks from disk back to cache.

This will happen in the case, there has been unclean cache shutdown.

However what is unclean is - why it slows down boot by hours.
Is the cache too big??

Regards


Zdenek

--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel

Re: dm-cache issue

[Teodor Milkov]

On 16.11.2016 11:24, Zdenek Kabelac wrote:
> Dne 15.11.2016 v 13:38 Teodor Milkov napsal(a):
>> On 14.11.2016 17:34, Zdenek Kabelac wrote:
>>> Dne 14.11.2016 v 16:02 Alexander Pashaliyski napsal(a):
>>>> The server is booting for hours, because of IO load. It seems is 
>>>> triggered
>>>> a flush from SSD disk (that is used for a cache device) to the raid
>>>> controllers (they are with slow SATA disks).
>>>> I have 10 cached logical volumes in *writethrough mode*, each with 
>>>> 2T of
>>>> data over 2 raid controllers. I use a single SSD disk for the cache.
>>>> The backup system is with lvm2-2.02.164-1 & kernel 4.4.30.
>>>>
>>>> Do you have any ideas why such flush is triggered? In writethrough 
>>>> cache mode
>>>> we shouldn't have dirty blocks in the cache.
>>>>
>>>
>>> Have you ensured there was proper shutdown ?
>>> Cache needs to be properly deactivated - if it's just turned off,
>>> all metadata are marked dirty.
>>>
>>> Zdenek
>>
>> Hi,
>>
>> I'm seeing the same behavior described by Alexander. Even if we assume
>> something is wrong with my shutdown scripts, still how could dm-cache 
>> ever be
>> dirty in writethrough mode? What about the case where server crashes for
>> whatever reason (kernel bug, power outage, operator error etc.)? Waiting
>> several hours, or for sufficiently large cache even days for the 
>> system to
>> come back up is not practical.
>>
>> I found this 2013 conversation, where Heinz Mauelshagen <heinzm 
>> redhat com>
>> states that "in writethrough mode the cache will always be coherent 
>> after a
>> crash": https://www.redhat.com/archives/dm-devel/2013-July/msg00117.html
>>
>> I'm thinking for a way to --uncache and recreate cache devices on 
>> every boot,
>> which should be safe in writethrough mode and takes reasonable, and more
>> importantly – constant amount of time.
>
> My first 'guess' in this reported case is - the disk I/O traffic seen 
> is related to the 'reload' of cached chunks from disk back to cache.
>
> This will happen in the case, there has been unclean cache shutdown.
>
> However what is unclean is - why it slows down boot by hours.
> Is the cache too big??

Indeed, cache is quite big – a 800GB SSD, but I found experimentally 
that this is the size where I get good cache hit ratios with my >10TB 
data volume.

As to the 'reload' vs 'flush' – I think it is flushing, because iirc 
iostat showed lots of SSD reading and HDD writing, but I'm not really 
sure and need to confirm that.

So, are you saying that in case of unclean shutdown this 'reload' is 
inevitable?

How much time it takes obviously depends on the SSD size/speed & HDD 
speed, but with 800GB SSD it is reasonable to expect very long boot times.

 > Can you provide full logs from 'deactivation' and following activation?

Any hints as to how to collect "full logs from 'deactivation' and 
following activation"? It happens early in the Debian boot process (I 
think udev does the activation) and I'm not sure how to enable 
logging... should I tweak /etc/lvm/lvm.conf?

Best regards,
Teodor Milkov

--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel

Re: dm-cache issue

[Zdenek Kabelac]

Dne 16.11.2016 v 14:45 Teodor Milkov napsal(a):
> On 16.11.2016 11:24, Zdenek Kabelac wrote:
>> Dne 15.11.2016 v 13:38 Teodor Milkov napsal(a):
>>> On 14.11.2016 17:34, Zdenek Kabelac wrote:
>>>> Dne 14.11.2016 v 16:02 Alexander Pashaliyski napsal(a):
>>>>> The server is booting for hours, because of IO load. It seems is triggered
>>>>> a flush from SSD disk (that is used for a cache device) to the raid
>>>>> controllers (they are with slow SATA disks).
>>>>> I have 10 cached logical volumes in *writethrough mode*, each with 2T of
>>>>> data over 2 raid controllers. I use a single SSD disk for the cache.
>>>>> The backup system is with lvm2-2.02.164-1 & kernel 4.4.30.
>>>>>
>>>>> Do you have any ideas why such flush is triggered? In writethrough cache
>>>>> mode
>>>>> we shouldn't have dirty blocks in the cache.
>>>>>
>>>>
>>>> Have you ensured there was proper shutdown ?
>>>> Cache needs to be properly deactivated - if it's just turned off,
>>>> all metadata are marked dirty.
>>>>
>>>> Zdenek
>>>
>>> Hi,
>>>
>>> I'm seeing the same behavior described by Alexander. Even if we assume
>>> something is wrong with my shutdown scripts, still how could dm-cache ever be
>>> dirty in writethrough mode? What about the case where server crashes for
>>> whatever reason (kernel bug, power outage, operator error etc.)? Waiting
>>> several hours, or for sufficiently large cache even days for the system to
>>> come back up is not practical.
>>>
>>> I found this 2013 conversation, where Heinz Mauelshagen <heinzm redhat com>
>>> states that "in writethrough mode the cache will always be coherent after a
>>> crash": https://www.redhat.com/archives/dm-devel/2013-July/msg00117.html
>>>
>>> I'm thinking for a way to --uncache and recreate cache devices on every boot,
>>> which should be safe in writethrough mode and takes reasonable, and more
>>> importantly – constant amount of time.
>>
>> My first 'guess' in this reported case is - the disk I/O traffic seen is
>> related to the 'reload' of cached chunks from disk back to cache.
>>
>> This will happen in the case, there has been unclean cache shutdown.
>>
>> However what is unclean is - why it slows down boot by hours.
>> Is the cache too big??
>
> Indeed, cache is quite big – a 800GB SSD, but I found experimentally that this
> is the size where I get good cache hit ratios with my >10TB data volume.

Yep - that's the current trouble of existing  dm-cache target.
It's getting inefficient when maintaining more then 1 million
cache block entries - recent versions of lvm2 even do not allow
create such cache without enforcing it.
(so for 32k blocks it'  ~30G cache data size)


> As to the 'reload' vs 'flush' – I think it is flushing, because iirc iostat
> showed lots of SSD reading and HDD writing, but I'm not really sure and need
> to confirm that.
>
> So, are you saying that in case of unclean shutdown this 'reload' is inevitable?

Yes - clean shutdown is mandatory - otherwise cache can't know consitency
and has to refresh itself.  Other option would be probably to drop cache
and let it rebuild - but you lose already gained 'knowledge' this way.

Anyway AFAIK there is ongoing devel and up-streaming process for new cache 
target which will others couple shortcomings and should perform much
better.   lvm2 will supposedly handle transition to a new format in some way
later.


> How much time it takes obviously depends on the SSD size/speed & HDD speed,
> but with 800GB SSD it is reasonable to expect very long boot times.
>
>> Can you provide full logs from 'deactivation' and following activation?
>
> Any hints as to how to collect "full logs from 'deactivation' and following
> activation"? It happens early in the Debian boot process (I think udev does
> the activation) and I'm not sure how to enable logging... should I tweak
> /etc/lvm/lvm.conf?


All you need to collect is basically 'serial' console log from your
machine  - so if you have other box to trap serial console log - it's
the most easiest option.

But since you already said you use  ~30times bigger cache size then the size 
with 'reasonable' performance - I think it's already clear where is your
problem hidden.

Until new target will be deployed - please consider to use significantly 
smaller cache size so the number of cache chunks is not above 1 000 000.

Regards

Zdenek




--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel

Re: dm-cache issue

[Teodor Milkov]

On 16.11.2016 16:06, Zdenek Kabelac wrote:
> Dne 16.11.2016 v 14:45 Teodor Milkov napsal(a):
>> On 16.11.2016 11:24, Zdenek Kabelac wrote:
>>> My first 'guess' in this reported case is - the disk I/O traffic 
>>> seen is
>>> related to the 'reload' of cached chunks from disk back to cache.
>>>
>>> This will happen in the case, there has been unclean cache shutdown.
>>>
>>> However what is unclean is - why it slows down boot by hours.
>>> Is the cache too big??
>>
>> Indeed, cache is quite big – a 800GB SSD, but I found experimentally 
>> that this
>> is the size where I get good cache hit ratios with my >10TB data volume.
>
> Yep - that's the current trouble of existing  dm-cache target.
> It's getting inefficient when maintaining more then 1 million
> cache block entries - recent versions of lvm2 even do not allow
> create such cache without enforcing it.
> (so for 32k blocks it'  ~30G cache data size)

I'm sorry for not being clear: similarly to the OP my SSD is split among 
10 LVs, so eache cache is around 80GB.

>> As to the 'reload' vs 'flush' – I think it is flushing, because iirc 
>> iostat
>> showed lots of SSD reading and HDD writing, but I'm not really sure 
>> and need
>> to confirm that.
>>
>> So, are you saying that in case of unclean shutdown this 'reload' is 
>> inevitable?
>
> Yes - clean shutdown is mandatory - otherwise cache can't know consitency
> and has to refresh itself.  Other option would be probably to drop cache
> and let it rebuild - but you lose already gained 'knowledge' this way.
>
> Anyway AFAIK there is ongoing devel and up-streaming process for new 
> cache target which will others couple shortcomings and should perform 
> much
> better.   lvm2 will supposedly handle transition to a new format in 
> some way
> later.
>
>> How much time it takes obviously depends on the SSD size/speed & HDD 
>> speed,
>> but with 800GB SSD it is reasonable to expect very long boot times.
>>
>>> Can you provide full logs from 'deactivation' and following activation?
>>
>> Any hints as to how to collect "full logs from 'deactivation' and 
>> following
>> activation"? It happens early in the Debian boot process (I think 
>> udev does
>> the activation) and I'm not sure how to enable logging... should I tweak
>> /etc/lvm/lvm.conf?
>
> All you need to collect is basically 'serial' console log from your
> machine  - so if you have other box to trap serial console log - it's
> the most easiest option.
>
> But since you already said you use  ~30times bigger cache size then 
> the size with 'reasonable' performance - I think it's already clear 
> where is your
> problem hidden.
>
> Until new target will be deployed - please consider to use 
> significantly smaller cache size so the number of cache chunks is not 
> above 1 000 000.

Thank you very much for your help! I'll give it another go at debugging 
what the problem is.
I found dm-writeboost in write_around_mode (kinda write-through) works 
well for me, so if I don't manage to get along with dm-cache I have plan B.


Best regards,
Teodor

--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel

Re: dm-cache issue

[Teodor Milkov]

[-- Attachment #1.1: Type: text/plain, Size: 1435 bytes --]

So, I ended up with the following setup:

  * Modified /etc/init.d/lvm2 to explicitly deactivate logical volumes:
    /lvm lvchange -a n --select name=~.*/
      o Create////lvm2-clean-shutdown /if deactivation is successful.//
  * Turn off automatic LV activation by
    setting///auto_activation_volume_list=[]/ in //etc/lvm/lvm.conf/
  * Try to activate LVs later in the boot only
    if//////lvm2-clean-shutdown /exists.//

This way I have the chance to try manual activation of the volumes in 
the case of unclean shutdown, or eventually uncache and recreate caches 
if flush/refresh appears to take too much time.

One strange thing is that while testing I forgot to set 
/auto_activation_volume_list=[] /and//boot process got stuck at an very 
early stage – probably at udev activating lvm //(////sbin/lvm pvscan 
--cache --activate ay/ from //lib/udev/rules.d/69-lvm-metad.rules)./// I 
waited for about five minutes but it didn't look like it's going to 
finish anytime soon, and I had no tty (too early in the boot), so I did 
reboot with a rescue CD and turned off LV auto activation in lvm.conf. 
This time the system came up quickly and the odd thing is that manual 
/lvchange -ay/ was pretty fast - less that a minute for all 12 × 60GB 
caches. So, maybe there's a bug in /udev/pvscan --activate/?

Anyway, the above setup seems to work well for me, so I'm sticking with 
it for the time being.

Best regards,
Teodor


[-- Attachment #1.2: Type: text/html, Size: 1979 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]