Closed kript closed 7 years ago
Interesting, I would have expected to see...
"rebalance for [%s] - updating object count from [%d] to [%d]",
from https://github.com/irods/irods/blob/4-1-stable/iRODS/server/core/src/miscServerFunct.cpp#L3215
@trel that function, update_resource_object_count()
, is used to update the cached object count of a resource (that is stored in r_resc_main
) with a value recomputed with a genquery. That is fixing a different type of situation than data object replication.
@kript as of 4.1.10, it is not expected that a rebalance
operation on a replication
resource would log the replication of data objects. The only way to know at this point that the rebalance
has done anything is through additional queries, e.g. your ils -l
.
Would you want logging added to the replication
resource for this?
We would like a way to know what is being done, yes. This is related to #3409
For example, we currently have ~24k single objects with some or all of them in one of the two composite resources we have in a zone. Because it is hard to get the information about which single object is in which resource tree, and there is no feedback from the rebalance command we don't know how long the command has to run, or even if its doing anything! :-)
Pulling this back into 4.1.11.
The logging will look something like this example:
proc_results_for_rebalance: rebalancing data id [10165] from [demoResc;unix3Resc] to [demoResc;unix2Resc]
in 4.1.10, the output doesn't appear to show when replicas are updated in a
rebalance
operation.I ran the following;
So the rebalance command did what it said it would do.
In the logs, however, there is no mention of having fixed a replication (apologies for the verbosity, we have a lot of dummy resources in dev to match production);
Should I see something? How would I know rebalance is doing its thing?