Skip re-compressing during startup_walk. This happens when restarting pghoard while the deletion queue has not been emptied. Meaning, that pghoard might had already compressed some files but not deleted them yet, which ends up on duplicated upload events.
For example during startup, WAL file 00001 exists in both /xlog and /xlog_incoming, pghoard will attempt to compress all files in /xlog_incoming and later upload files in /xlog. The transfer agent will end up adding 2 upload events for 00001 compressed file, but the second event will fail since the compressed file might have been already deleted, so the transfer agent will keep retrying and end up getting stuck.
Missing uploading compressed timeline files, noticed that during the startup walk, pghoard is only checking files in xlog path, but compressed timelines are stored under /timelines.
About this change - What it does
Skip re-compressing during startup_walk. This happens when restarting pghoard while the deletion queue has not been emptied. Meaning, that pghoard might had already compressed some files but not deleted them yet, which ends up on duplicated upload events.
For example during startup, WAL file
00001
exists in both/xlog
and/xlog_incoming
, pghoard will attempt to compress all files in/xlog_incoming
and later upload files in/xlog
. The transfer agent will end up adding 2 upload events for00001
compressed file, but the second event will fail since the compressed file might have been already deleted, so the transfer agent will keep retrying and end up getting stuck.Missing uploading compressed timeline files, noticed that during the startup walk, pghoard is only checking files in
xlog
path, but compressed timelines are stored under/timelines
.Resolves: #BF-1358