Log in New API Help About

Dave - Question time for Allan

1.2 KB of Plain text
Created 7 months ago — expires in 150 days
This one's for Allan!

I have a small zpool that has been `zfs send -LevcpR > /tmp/stash.zfs` and the stash.zfs moved to a distribution server. The zpool has ~ 3000 datasets + clones. The initial send with most of the data was quick as usual, but the tail end where the clones are being received and linked up to the parent dataset, seems *very* slow:

each attaching the clone and receiving the incremental stream takes ~ 15 seconds, even though data is in most cases a few Kib.

any ideas why, and if it's possible to make this happen faster?

... 15 seconds pass
found clone origin zroot/koans/templates/12.1-RELEASE@pristine
receiving incremental stream of zroot/koans/clones/h2o/c80073e5e1c5ad5effc4e858fb8f5d5fb9b61e93@pristine into zroot/koans/clones/h2o/c80073e5e1c5ad5effc4e858fb8f5d5fb9b61e93@pristine
received 624B stream in 23 seconds (27B/sec)

... 15 seconds pass
receiving incremental stream of zroot/koans/clones/h2o/c80073e5e1c5ad5effc4e858fb8f5d5fb9b61e93@pre-debug into zroot/koans/clones/h2o/c80073e5e1c5ad5effc4e858fb8f5d5fb9b61e93@pre-debug
received 312B stream in 1 seconds (312B/sec)
... 15 seconds pass

Most of the datasets are base jail template clones.