insufficient buffers to rebalance the tree

Pierre Abbat phma at phma.optus.nu
Mon Mar 15 02:16:46 PDT 2010


On Monday 15 March 2010 00:06:41 Matthew Dillon wrote:
>     The hammer rebalance code isn't critical to filesystem operation
>     but it needs a lot of buffer cache buffers to operate, and systems
>     running in VMs with small amounts of ram usually do not have enough
>     buffers to cover the needs.
>
>     You should be able to ignore the issue inside a VM.

I'm not running in a VM. There is no free command, so here's the memory line 
from top:

Memory: 53M Active, 258M Inact, 171M Wired, 10M Cache, 57M Buf, 1596K Free

Attempting to run "hammer rebalance" directly results in this:

dogla# hammer rebalance /
rebalance start 8000000000000000:0000
Rebalance / failed: No space left on device
Rebalance:
    0 btree nodes scanned
    0 btree nodes deleted
    0 collision retries
    0 btree nodes rebalanced

I've tried /dev/ad0s1a, /pfs/var, and ROOT for the filesystem; none of them 
work.

Pierre
-- 
Jews use a lunisolar calendar; Muslims use a solely lunar calendar.





More information about the Users mailing list