Re: [HACKERS] Backend Cache Invalidation - Mailing list pgsql-hackers

From Tom Lane
Subject Re: [HACKERS] Backend Cache Invalidation
Date
Msg-id 14025.929371226@sss.pgh.pa.us
Whole thread Raw
In response to Backend Cache Invalidation  ("Tim Perdue" <perdue@raccoon.com>)
List pgsql-hackers
"Tim Perdue" <perdue@raccoon.com> writes:
> Is anyone willing to tell me what this means:
> NOTICE:  SIAssignBackendId: discarding tag 2147449930
> Connection to database 'db_domain' failed.
> FATAL 1:  Backend cache invalidation initialization failed

This is coming from SIAssignBackendId() in
backend/storage/ipc/sinvaladt.c.  Maybe I'm missing something,
but the logic in that routine (as of current CVS sources)
sure looks wrong.

Lines 130-138 look like they are checking for no-longer-used
procState entries that can be recycled, but (a) the test is
not obviously correct (and I'm always suspicious of code that's
both this complicated and comment-free); (b) lines 151-158
reject any entry that doesn't have tag InvalidBackendTag or
backendTag, meaning that lines 130-138 could be diked out with
no loss of functionality.

Something is wrong here...
        regards, tom lane


pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Re: [HACKERS] regress.sh
Next
From: Thomas Lockhart
Date:
Subject: Re: [HACKERS] regress.sh