[linux-lvm] lvm cache on debian 8, missing cache information

[Manuel Bernal Llinares]

Hi guys,

I’m trying to set up lvmcache for a backup volume I use for time machine.

I’ve seen there is quite a lot of very good documentation about this, but also out of sync between different linux systems, e.g. I can’t get those cache* attributes when running lvs on debian 8.

As you can see under this lines, “tmachine” is a cached LVM volume
# lvs -a
LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert
backups blackhole -wi-ao---- 5,00t
clasificador blackhole -wi-ao---- 1,00t
[lvol0_pmspare] blackhole ewi------- 592,00m
nfs_home blackhole -wi-ao---- 1,00t
tmachine blackhole Cwi-aoC--- 3,00t tmachine_cache [tmachine_corig]
tmachine_cache blackhole Cwi---C--- 231,67g
[tmachine_cache_cdata] blackhole Cwi-ao---- 231,67g
[tmachine_cache_cmeta] blackhole ewi-ao---- 592,00m
[tmachine_corig] blackhole owi-aoC--- 3,00t

Even with a couple of mac computers sending its time machine data for a while, the cache seems to not be working.

This are the parameters described by “dmsetup status” when run on that volume:

# dmsetup status /dev/blackhole/tmachine
0 6442450944 cache 8 12849/151552 128 11886/3795648 17772 177033 2798405 1620326 0 3856 1 1 writeback 2 migration_threshold 2048 mq 10 random_threshold 4 sequential_threshold 512 discard_promote_adjustment 1 read_promote_adjustment 4 write_promote_adjustment 8

I’ve tested another cache alternative (flashcache) before, and the parameter "sequential_threshold” sounds familiar to me, I suspect it is the one making the cache flag the data transfer as “sequential”, passing it through to the backend (slow) device, instead of using the cache to absorb the writing traffic, and clean dirty blocks later, behavior I saw when using flashcache.

“sequential_threshold” is configured, by default, with a value of 512, but I haven’t been able to find what’s that 512, maybe extents, bytes, kb, mb? Unfortunately, I haven’t been able to find a way, at least in Debian, to change those parameters, neither creating the cache nor once it’s already been created.

Any ideas?

Thank you for your help guys!!!





--------
Manuel Bernal Llinares
***@gmail.com
M.Sc. in Computer Science

[Zdenek Kabelac]

Post by Manuel Bernal Llinares
Hi guys,
I’m trying to set up lvmcache for a backup volume I use for time machine.
I’ve seen there is quite a lot of very good documentation about this, but also out of sync between different linux systems, e.g. I can’t get those cache* attributes when running lvs on debian 8.
As you can see under this lines, “tmachine” is a cached LVM volume
# lvs -a
LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert
backups blackhole -wi-ao---- 5,00t
clasificador blackhole -wi-ao---- 1,00t
[lvol0_pmspare] blackhole ewi------- 592,00m
nfs_home blackhole -wi-ao---- 1,00t
tmachine blackhole Cwi-aoC--- 3,00t tmachine_cache [tmachine_corig]
tmachine_cache blackhole Cwi---C--- 231,67g
[tmachine_cache_cdata] blackhole Cwi-ao---- 231,67g
[tmachine_cache_cmeta] blackhole ewi-ao---- 592,00m
[tmachine_corig] blackhole owi-aoC--- 3,00t
Even with a couple of mac computers sending its time machine data for a while, the cache seems to not be working.
# dmsetup status /dev/blackhole/tmachine
0 6442450944 cache 8 12849/151552 128 11886/3795648 17772 177033 2798405 1620326 0 3856 1 1 writeback 2 migration_threshold 2048 mq 10 random_threshold 4 sequential_threshold 512 discard_promote_adjustment 1 read_promote_adjustment 4 write_promote_adjustment 8
I’ve tested another cache alternative (flashcache) before, and the parameter "sequential_threshold” sounds familiar to me, I suspect it is the one making the cache flag the data transfer as “sequential”, passing it through to the backend (slow) device, instead of using the cache to absorb the writing traffic, and clean dirty blocks later, behavior I saw when using flashcache.
“sequential_threshold” is configured, by default, with a value of 512, but I haven’t been able to find what’s that 512, maybe extents, bytes, kb, mb? Unfortunately, I haven’t been able to find a way, at least in Debian, to change those parameters, neither creating the cache nor once it’s already been created.
Any ideas?

'mq' policy is rather obsoleted.

Now it's suggested to use 'smq' which should be better in all aspects.
So please try usage of new 'smq' first (which happens to have no configurables
ATM - it's auto-tunning and consumes less memory).

You also need relatively new lvm2 to show all cache settings via lvm2 'lvs' attrs.

Regards

Zdenek

[Manuel Bernal Llinares]

Thanks for the reply Zdenek!

I use LVM2 from apt repo for debian
ii liblvm2cmd2.02:amd64 2.02.111-2.2 amd64 LVM2 command library
ii lvm2 2.02.111-2.2 amd64 Linux Logical Volume Manager

And, for creating the cache volume, I used lvcreate to do it in a single command, something like this

lvcreate -n tmachine_cache --type cache-pool --cachemode writeback -l 100%FREE blackhole <fast_device_to_use_as_cache>

And… I just had a look at the redhat repository and… yeah, this version of LVM2 running on Debian seems to be dated back on September 2014.

Is this version of LVM2 too old for using lvmcache?

Thanks! ^_^

Post by Zdenek Kabelac

Post by Manuel Bernal Llinares
Hi guys,
I’m trying to set up lvmcache for a backup volume I use for time machine.
I’ve seen there is quite a lot of very good documentation about this, but also out of sync between different linux systems, e.g. I can’t get those cache* attributes when running lvs on debian 8.
As you can see under this lines, “tmachine” is a cached LVM volume
# lvs -a
LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert
backups blackhole -wi-ao---- 5,00t
clasificador blackhole -wi-ao---- 1,00t
[lvol0_pmspare] blackhole ewi------- 592,00m
nfs_home blackhole -wi-ao---- 1,00t
tmachine blackhole Cwi-aoC--- 3,00t tmachine_cache [tmachine_corig]
tmachine_cache blackhole Cwi---C--- 231,67g
[tmachine_cache_cdata] blackhole Cwi-ao---- 231,67g
[tmachine_cache_cmeta] blackhole ewi-ao---- 592,00m
[tmachine_corig] blackhole owi-aoC--- 3,00t
Even with a couple of mac computers sending its time machine data for a while, the cache seems to not be working.
# dmsetup status /dev/blackhole/tmachine
0 6442450944 cache 8 12849/151552 128 11886/3795648 17772 177033 2798405 1620326 0 3856 1 1 writeback 2 migration_threshold 2048 mq 10 random_threshold 4 sequential_threshold 512 discard_promote_adjustment 1 read_promote_adjustment 4 write_promote_adjustment 8
I’ve tested another cache alternative (flashcache) before, and the parameter "sequential_threshold” sounds familiar to me, I suspect it is the one making the cache flag the data transfer as “sequential”, passing it through to the backend (slow) device, instead of using the cache to absorb the writing traffic, and clean dirty blocks later, behavior I saw when using flashcache.
“sequential_threshold” is configured, by default, with a value of 512, but I haven’t been able to find what’s that 512, maybe extents, bytes, kb, mb? Unfortunately, I haven’t been able to find a way, at least in Debian, to change those parameters, neither creating the cache nor once it’s already been created.
Any ideas?

'mq' policy is rather obsoleted.
Now it's suggested to use 'smq' which should be better in all aspects.
So please try usage of new 'smq' first (which happens to have no configurables ATM - it's auto-tunning and consumes less memory).
You also need relatively new lvm2 to show all cache settings via lvm2 'lvs' attrs.
Regards
Zdenek
_______________________________________________
linux-lvm mailing list
https://www.redhat.com/mailman/listinfo/linux-lvm
read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/

--------
Manuel Bernal Llinares
***@gmail.com
M.Sc. in Computer Science