I have multiple LVs in a VG:
LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert
cyrus_spool centos Vwi-aotz-- 29.80g pool00 95.84
cyrus_spool.old centos Vwi-a-tz-- 22.18g pool00 98.91
data centos Vwi-aotz-- 25.00g pool00 77.88
debuginfo centos -wi-ao---- 4.73g
home centos Vwi-aotz-- 10.00g pool00 86.77
home.old centos Vwi-a-tz-- 10.00g pool00 99.92
http_cache centos Vwi-aotz-- 50.00g pool00 72.10
mp3 centos Vwi-aotz-- 150.00g pool00 84.37
nextcloud_data centos Vwi-aotz-- 250.00g pool00 71.25
photos centos -wi-ao---- 64.49g
pkg-cache centos Vwi-aotz-- 47.68g pool00 81.41
pool00 centos twi-aotz-- 1.26t 75.81 35.35
root centos Vwi-aotz-- 28.81g pool00 16.68
snaptest centos Vwi---tz-k 1.46g pool00 root
source centos Vwi-aotz-- 92.00g pool00 92.31
swap centos Vwi-aotz-- 8.00g pool00 100.00
swap2 centos -wi-a----- 16.00g
test centos -wi-a----- 100.00m
usr centos Vwi-aotz-- 9.25g pool00 77.97
var centos Vwi-aotz-- 265.45g pool00 92.25
windows10 centos -wi-a----- 6.00g
that I want to cache with a faster 120GB SSD. Not all of those LVs are
even actively used, and some are rarely used, but I would think any
decent caching algorithm would work out the most used LVs and their
hotspots so trying to cherry-pick which ones to cache and which not to
shouldn't even be an issue.
But as I understand it, I need a separate caching LV (for cachevol) or
two (for cachepool) for each LV that I want to cache.
This seems rather sub-optimal in trying to "guess" what portion of the
120GB SSD I should use for each LV that I want to cache, and trying to
guess which LVs would be most [in]effective to cache even.
Is there no better way to do this than 1:1 cache{pool|vol} per LV?
I suppose I could take a WAG and then after a while evaluate the stats
and decide (manually) if some re-balancing of the portioning would be
beneficial, and then keep doing that periodically but that seems rather
tedious and manual. Any solutions here to automate this, as a
workaround to this 1:1 LV requirement?
Alternatively, if it must be 1:1 would attaching the cache to the thin-
pool device, pool00 effect what I am trying to achieve? Is that even
possible?
Is lvmcache maybe just not the right solution in this case?
Cheers,
b.
