rrhodes@firstenergycorp.com wrote:
I'm thinking of running a “reallocate start –p” on a couple volumes that are showing high latency disk accesses for luns. The volume is a source for both snapmirror and snapvault operations.
Right, you do have active snapshots (since you're mentioning SnapVault below) I take it. reallocate start -f -p <vol_name> Are you seeing this high latency for READ I/Os above all on these LUNs / Vols? If it's for WRITE, your issue might be a different one and your efforts running that command won't do as much good. Could still be layout related, Free Space Fragmentation e.g.
(I'm not at all sure what you mean by "disk accesses" in this context.)
I was reading up on this command and I can’t find any real info on the effect of running this on the snapmirror and snapvault operations.
Question: If I run a “reallocate start –p” on a volume, will snapmirror/snapvault see this and send all reallocated blocks? Also, can snapmirror/snapvault operations be running during the reallocate?
I'm not 100% certain (yet), but incidentally I'm digging around in this area as well since a while now, effects on SnapVault and the snapshots it uses on the source. And as you say, the info is rather frugal.
To my current knowledge, the answer to your first Q is: no. However there will, during a certain time, be extra directs inside WAFL when blocks from snapshots are read and that might slow things down. No idea how much :-( This is automatically corrected by the reallocate -p, which will start wafl reallocate scan in the background that cleans it up for the FlexVol in question. It looks like this (example):
$ xyz 'priv set -q diag; wafl scan status' | egrep redirect | sed 's/ //' 138932 redirect public phase 1 level 2 inode 5970287 of 33554409 (2%) 138942 redirect private phase 3 level 1 block 470560972 of 771868448 (22%) 139000 redirect public phase 1 level 1 inode 19865752 of 33554409 (1%) 139884 redirect public phase 1 level 1 inode 24542520 of 33554409 (3%) 139054 redirect public phase 3 level 1 block 349198358 of 410692272 (33%) 140078 redirect public phase 1 level 2 inode 32509281 of 33554409 (7%) 140519 redirect public phase 1 level 1 inode 11041527 of 113246091 (0%) 140427 redirect public phase 1 level 1 inode 2218329 of 33554409 (0%) 139956 redirect public phase 1 level 1 inode 11865625 of 33554409 (1%) 140125 redirect public phase 1 level 1 inode 17613080 of 33554409 (1%) 138844 redirect private phase 3 level 1 block 305546101 of 537103936 (21%) 138991 redirect public phase 1 level 2 inode 19646076 of 33554409 (2%) 140192 redirect public phase 1 level 1 inode 24762423 of 33554409 (3%) 139853 redirect public phase 3 level 1 block 61740937 of 258940032 (14%) 140521 redirect public phase 2 level 1 block 2279190 of 7518795 (10%)
During this scan, the FlexVol continues to operate normally in any other respect so to speak. You can even abort such a scan cleanly and have it implicitly restarted later, with no harm to anything in the FlexVol.
To the second Q: yes
/M