Database error in transaction not recovered during bulk loading
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
hockeypuck |
Fix Committed
|
High
|
Casey Marshall |
Bug Description
<SysComet> "current transaction is aborted, commands ignored until end of transaction block", thus I have errors but not a clue as to the key which caused it.
<cmars> was it spewing from the start like this, or running for a while and then failed?
<SysComet> Looks like it was while loading, from a dump from keyserver.
<SysComet> Running fine for the first few .pgp files it loaded, then when I came back to it an hour or so later, I saw this.
<SysComet> So, if I just drop the tables from that schema, I should be good to just reload, right? Oh, and nuke the ptree dir
<cmars> yes
<SysComet> hkp=> select count(*) from openpgp_pubkey;
<SysComet> -[ RECORD 1 ]-
<SysComet> count | 245000
Need to recover with a rollback on errors like this. Would be nice to understand the cause of the error. Too many missed keys like this could pose a problem for converging reconciliation with SKS.
Best to stop loading on database error immediately, report the issue. A bulk load can always be resumed. Database errors are NOT normal during a bulk load, and indicate something is wrong.