From: | Fujii Masao <masao(dot)fujii(at)oss(dot)nttdata(dot)com> |
---|---|
To: | Bharath Rupireddy <bharath(dot)rupireddyforpostgres(at)gmail(dot)com> |
Cc: | PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
Subject: | Re: Inconsistency in startup process's MyBackendId and procsignal array registration with ProcSignalInit() |
Date: | 2021-10-11 15:29:03 |
Message-ID: | 2222ab6f-46b1-d5c0-603d-8f6680739db4@oss.nttdata.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
On 2021/10/11 19:46, Bharath Rupireddy wrote:
> If we do the above, then the problem might arise if somebody calls
> SICleanupQueue and wants to signal the startup process, the below code
> (from SICleanupQueue) can't get the startup process backend id. So,
> the backend id calculation for the startup process can't just be
> MaxBackends + MyAuxProcType + 1.
> BackendId his_backendId = (needSig - &segP->procState[0]) + 1;
Attached POC patch illustrates what I'm in mind. ISTM this change
doesn't prevent SICleanupQueue() from getting right backend ID
of the startup process. Thought?
> It looks like we need to increase the size of the ProcState array by 1
> at least (for the startup process). Currently the ProcState array
> doesn't have entries for auxiliary processes, it does have entries for
> MaxBackends. The startup process is eating up one slot from
> MaxBackends. Since we need only an extra ProcState array slot for the
> startup process I think we could just extend its size by 1. Instead of
> modifying the MaxBackends definition, we can just add 1 (and a comment
> saying this 1 is for startup process) to shmInvalBuffer->maxBackends
> in SInvalShmemSize, CreateSharedInvalidationState. IMO, this has to go
> in a separate patch and probably in a separate thread. Thoughts?
Agreed.
Regards,
--
Fujii Masao
Advanced Computing Technology Center
Research and Development Headquarters
NTT DATA CORPORATION
Attachment | Content-Type | Size |
---|---|---|
startup_backendid.patch | text/plain | 2.3 KB |
From | Date | Subject | |
---|---|---|---|
Next Message | David G. Johnston | 2021-10-11 15:44:20 | Re: Proposal: allow database-specific role memberships |
Previous Message | PG Bug reporting form | 2021-10-11 15:25:49 | BUG #17220: ALTER INDEX ALTER COLUMN SET (..) with an optionless opclass makes index and table unusable |