A small revisit of this, I'd like to show y'all something.
John Clear wrote:
As others have mentioned, you want to let deswizzling finish. You can tune the priority of the deswizzling process with 'wafl scan speed' to limit the impact of the scan.
There's a vol in a 8060 of ours, it has one (1) large Aggr with FlashPool.
$ naXXX version NetApp Release 8.2.3P2D10 7-Mode: Thu Apr 9 22:11:11 PDT 2015 $ naXXX 'priv set -q diag; wafl scan speed' # Auto tuning is active WAFL scan speed is 6800 $ naXXX 'priv set -q diag; wafl scan status' | egrep 'volp09201|deswizz' Volume volp09201: 25 volume deswizzling snap 67, inode 24342696 of 157894716. level 1 of normal files. Totals: Normal files: L1:1/216489640 L2:0/132059634 L3:0/39636101 L4:0/0 Inode file: L0:0/0 L1:0/0 L2:0/0 L3:0/0 L4:0/0
The da** thing walks inodes, and in all snapshots dirs as well *sigh*. It can also easily swamp the disks because it reads data fairly inefficiently.
This FlexVol was migrated over with VSM from a different system, which has been shut down now. The cut over change was >8 weeks ago. There is *probably* some way, might not be user accessible at all though (not even from a BSD shell inside the node), to crank up the prio of that particular scanner to make it run faster but still it will in practice never finish. BG DATA w.r.t. inodes.
I don't know if it has finished deswizzling the "live" data. I couldn't care less about the .snapshot data as it's only there for the end users to recover "self service" if they screw up. Unfortunately there's no way to make this scanner skip all the snapshots once it has fixed up all the "primary data", that would certainly help us here very very much.
(Yes I'm aware we're just silly here w.r.t. no of inodes and also NFS workload, but it is the way it is :-\ )
So time for a wee test, let's turn it off.
$ naXXX options wafl.deswizzle.enable off $ naXXX 'priv set -q diag; wafl scan status' | egrep 'deswizz'
Sure enough, that killed it. It's supposed to re-start again automagically on any vols in need of it once the option is put back to on.
$ naXXX options wafl.deswizzle.enable on $ naXXX 'priv set -q diag; wafl scan status' | egrep 'deswizz' 26924 volume deswizzling snap 67, inode 319722 of 157894716. level 1 of normal files. Totals: Normal files: L1:0/18487 L2:0/32712 L3:0/8326 L4:0/0 Inode file: L0:0/0 L1:0/0 L2:0/0 L3:0/0 L4:0/0
Yup. N.B. it started over from the beginning of the inode "tree" in snap #67 :-( :-(
Now, what I'd like to do is to make this volp09201 exempt from any more deswizzling. Other volumes might be AOK to do it on and would finish within reasonable time.
Alternatively, NetApp R&D could put a good resource on this to re-write it and optimize it well so that it's as clever about what needs to be done as possible. Until then in a HFC (High File Count) type of environment like ours, deswizzling is kind of moot.
I'll try and find out if a reallocate -A has an effect on what the deswizzler needs to do, and share my findings here later (after summer).
/M