Details
-
Bug
-
Resolution: Won't Fix
-
Major
-
None
-
7.1.3
-
Untriaged
-
0
-
Unknown
Description
My 50MB (uncompressed) dataset took over 500M of disk space despite having the default 50% fragmentation threshold specified. Only after fully stopping the workload for several minutes was the disk space reclaimed (bottomed out at around 160M).
Logs here: s3://cb-customers-secure/davidh/2023-02-03/collectinfo-2023-02-03t174442-ns_1@127.0.0.1.zip
Setup
- Single 7.1.3 node with one Magma bucket and 2 collections.
- Pillowfight command to populate 50M worth of data as 100,000 x 512 byte JSON docs at 200 ops / sec
cbc-pillowfight -U couchbase://10.145.233.101/charlie_magma --collection _default.incoming -m 512 -M 512 -J -p DH_ -I 100000 -t1 -r100 -u admin -P password --rate-limit 200
- Eventing function to move the data set from one collection to another:
function OnUpdate(doc, meta) {
dest[meta.id]=doc
delete source[meta.id]
}
Observations
The disk space used by magma far exceeded the total size of the data set, allowing the fragmentation to greatly exceed the specified level of 50%.
Only completely halting the workload allowed the disk space to come back under control.