The future of HAMMER1

[Daniel Bilik]

On Sun, 22 Jul 2018 11:20:06 -0700
Matthew Dillon <dillon at backplane.com> wrote:

> ...
> but single-storage support is rock solid and H2's compression features
> are really nice.

I've been using H2 on a desktop machine for several months now, and I must
confirm it works nicely. The only problem I've encountered so far is that
after some time the periodic cleanup starts to fail with error...

ioctl: Numerical argument out of domain

... with kernel output like this...

hammer2: bulkfree buf=  359M pass 0000000000000000-00000037b1800000 (11504GB of media)
 chains 100000  inodes 56475   dirents 73999   bytes   728MB
 chains 200000  inodes 149868  dirents 176228  bytes   898MB
 chains 300000  inodes 244094  dirents 292013  bytes  1060MB
 chains 400000  inodes 339717  dirents 383336  bytes  1196MB
 chains 500000  inodes 437082  dirents 463581  bytes  1325MB
 chains 600000  inodes 528924  dirents 578238  bytes  1469MB
chain 0000001d4251800e.02 key=0000000000022e00 meth=30 CHECK FAIL (flags=00144002, bref/data bf3e4c6ffbf5f50b/d644a08e4e408344)
chain 0000001d4251800e.02 key=0000000000022e00 meth=30 CHECK FAIL (flags=00144002, bref/data bf3e4c6ffbf5f50b/d644a08e4e408344)
 chains 700000  inodes 623088  dirents 674417  bytes  1609MB
 chains 800000  inodes 717729  dirents 759216  bytes  1755MB
 chains 900000  inodes 814501  dirents 871247  bytes  1888MB
bulkfree lastdrop 1 0 error=0x0002
bulkfree pass statistics (0.00% storage processed):
    bulkfree was aborted

I suspect this to be result of dirty unmounts (the machine has experienced
some hard lockups), or maybe it's somehow related to snapshots (I've
written a short shell script that maintains snapshots as a part of daily
cron).

Matt, do you have any idea how to fix this so that bulkfree can finish
successfully?

Thank you.

--
						Dan