Hi folks,
On our /scratch filesystem we have a range of users including some who
have *large* directories (I normally set alerting to anything over 25k
entries, but today I'm trying to scan one with 1.7 million files in
Robinhood was launched with a fresh clean DB (we're doing partial
scans/purges per project - until there's a way to limit purging to a
particular path on the commandline, this way seems to be fastest
lsof | grep /scratch on the robinhood server starts off normally (54K
directories under this path, but once it hits the huge one, all other
threads stop and the scan slows right down.
during this time the logging hasn't changed entries
scanned: 4819388 (or 'progress' line in the logfile) but the filecount
goes up
ie
aelwell@esDM012:~$ rbh-report -aif /ivec/etc/robinhood/scratch.conf ;
sleep 20 ; rbh-report -aif /ivec/etc/robinhood/scratch.conf
Filesystem scan activity:
Current scan interval: 180.0d
Last filesystem scan:
status: running
start: 2015/12/10 12:47:05 (5h 41min 25s ago)
last action: 2015/12/10 18:27:04 (01min 26s ago)
Statistics:
entries scanned: 4819388
errors: 0
timeouts: 0
# threads: 16
average speed: 2424.24 entries/sec
>>> current speed: 637.70 entries/sec
Storage usage has never been checked
No purge was performed on this filesystem
type , count, volume, avg_size
symlink , 1452, 91.16 KB, 64
dir , 54316, 548.62 MB, 10.34 KB
file , 5190003, 8.93 TB, 1.80 MB
Total: 5245771 entries, 9821173499141 bytes (8.93 TB)
Filesystem scan activity:
Current scan interval: 180.0d
Last filesystem scan:
status: running
start: 2015/12/10 12:47:05 (5h 41min 45s ago)
last action: 2015/12/10 18:27:04 (01min 46s ago)
Statistics:
entries scanned: 4819388
errors: 0
timeouts: 0
# threads: 16
average speed: 2424.24 entries/sec
>>> current speed: 637.70 entries/sec
Storage usage has never been checked
No purge was performed on this filesystem
type , count, volume, avg_size
symlink , 1452, 91.16 KB, 64
dir , 54316, 548.62 MB, 10.34 KB
file , 5190393, 8.93 TB, 1.80 MB
Total: 5246161 entries, 9821175974885 bytes (8.93 TB)
aelwell@esDM012:~$
which part of the scan cycle does the progress get updated?
I've also spotted another 'feature' (this is with 2.5.5) - if I start
robinhood with a command line like
robinhood --scan=/scratch/partner1065 --no-gc --once --purge --detach
-f /ivec/etc/robinhood/scratch.conf
and sigterm during scanning, it then goes on to start purging and
won't respond to another sigterm. Is this a bug or a feature? - I'm
guessing this is the logic buried in rbh_daemon.c in
if ( terminate_sig != 0 )
loop?
Andrew
------------------------------------------------------------------------------
_______________________________________________
robinhood-support mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/robinhood-support
