Feature #3137
closed
Write bytes read and network performance stats to arv-mount logs
Description
Toggle with a --crunchstat-interval argument (default 0 means don't output stats).
Use the same format as crunchstat, so the results show up in Workbench graphs. Example:
crunchstat: blkio:0:0 123456 write 1234560 read -- interval 10.0002 seconds 0 write 0 read crunchstat: net:keep0 1546380901 tx 3166208 rx -- interval 10.0002 seconds 1546352010 tx 2066527 rx
blkio:0:0 indicates filesystem traffic between userspace programs and arv-mount.
net:keep0 indicates network traffic between arv-mount and keep servers.
Additionally, it would be helpful to know the number of IO operations on both sides, i.e., number of Keep block reads/writes and number of userspace read() and write() calls serviced. (Formatting TBD?)
Updated by Peter Amstutz over 10 years ago
- Subject changed from arv-mount logs bytes read for stats to Write bytes read and network performance stats to arv-mount logs
Updated by Tom Clegg about 9 years ago
- Description updated (diff)
- Category set to FUSE
- Story points set to 1.0
Updated by Brett Smith about 9 years ago
Reviewing db55813
Heads up, arvados_fuse needs its dependency on arvados-python-client bumped to get arvados.keep.Counter.
Beyond that, I just have a couple of naming suggestions, nothing to block a merge:
- In the Stat class, the relationship between "egress" and "egr" is not clear from the names. Ditto "ingress" and "ing." Would better names help clarify that one is a specific value in time from the other?
Possibly related: there doesn't seem to be any need to hold onto "egr_prev" and "ing_prev" as instance variables. They only function as local variables to calculate the interval value in the update method. - In the Counter class, "lk" seems like a sort of gratuitous abbreviation of "lock".
Thanks.
Updated by Peter Amstutz about 9 years ago
- Status changed from New to Resolved
- % Done changed from 50 to 100
Applied in changeset arvados|commit:d6f163ed99c6ae234c551bb3ecd23c7af8610c8f.