Details
-
Bug
-
Status: Closed
-
Major
-
Resolution: Fixed
-
Cheshire-Cat
-
Untriaged
-
1
-
Yes
Description
Steps to reproduce:
1. Create a bucket which has been populated by data.
2. Create an archive and repository using the backup service which backs up to the cloud.
3. Take two backups (the first one was full followed by an incremental backup).
4. Obtain the 'info' either through the service or via cbbackupmgr.
Observe that the backup was completed successfully and a 1000 items were backed up.
The 'cbbackupmgr info' command is failing.
[root@node1-cheshire-cat-testing-centos7 data]# /opt/couchbase/bin//cbbackupmgr info --archive s3://c039dc20-0ec2-11eb-80c3-acde48001122/archive-10.112.210.101 --obj-access-key-id asdf --obj-endpoint http://10.112.202.104:4572 --obj-region asdf --obj-secret-access-key asdf --obj-staging-dir /tmp/altstaging --s3-force-path-style --json --all |
{"name":"archive-10.112.210.101","archive_uuid":"49a0fc93-b85f-425d-81cd-e12c6ac057e1","repos":[{"name":"4eee829d-df7e-40dd-8ba8-af8f477d4d07","size":1126,"count":2,"backups":[{"date":"2020-10-15T08_45_36.053096045Z","type":"FULL","source":"http://127.0.0.1:8091","events":0,"fts_alias":0,"size":563,"buckets":[{"name":"default","size":559,"items":0,"mutations":0,"tombstones":0,"views_count":0,"fts_count":0,"index_count":0,"analytics_count":0}],"complete":true,"source_cluster_uuid":"b42ca123505bb99db26cb66dd20806dd"},{"date":"2020-10-15T08_46_28.556572914Z","type":"INCR","source":"http://127.0.0.1:8091","events":0,"fts_alias":0,"size":563,"buckets":[{"name":"default","size":559,"items":0,"mutations":0,"tombstones":0,"views_count":0,"fts_count":0,"index_count":0,"analytics_count":0}],"complete":true,"source_cluster_uuid":"b42ca123505bb99db26cb66dd20806dd"}]}],"size":1126} |
The files in the data directory are not unzipped.
[root@node1-cheshire-cat-testing-centos7 data]# ls /tmp/altstaging/archive-10.112.210.101/4eee829d-df7e-40dd-8ba8-af8f477d4d07/2020-10-15T08_45_36.053096045Z/default-513add1ac41f543ac38bafe2932e9885/data/ |
failoverlogs.zip snapshots.zip stats.zip
|
The service also reports the same findings.
Asad Zaidi and I had a quick talk about this issue, it looks to be caused by the fact that when mounting an archive without locking we won't make changes to the archive itself (something changed recently; related to the staging directory revision id). The reason being, we don't want a read only command modifying the archive whilst another (read/write) command is actively using it. The fix will likely be to ensure that the process that populates the staging directory, always inflates the archive metadata (whilst still having a fallback for the case where users manually sync a cloud archive locally).