Re: logical replication busy-waiting on a lock

On 03/06/17 03:25, Andres Freund wrote:
>
>> That should solve the original problem reported here.
>
> Did you verify that?
>

Yes, I tried to manually create multiple exported logical decoding
snapshots in parallel and read data from them and it worked fine, while
it blocked before.

>
>> @@ -1741,17 +1741,17 @@ GetSerializableTransactionSnapshotInt(Snapshot snapshot,
>> } while (!sxact);
>>
>> /* Get the snapshot, or check that it's safe to use */
>> - if (!TransactionIdIsValid(sourcexid))
>> + if (!sourcevxid)
>> snapshot = GetSnapshotData(snapshot);
>> - else if (!ProcArrayInstallImportedXmin(snapshot->xmin, sourcexid))
>> + else if (!ProcArrayInstallImportedXmin(snapshot->xmin, sourcevxid))
>> {
>> ReleasePredXact(sxact);
>> LWLockRelease(SerializableXactHashLock);
>> ereport(ERROR,
>> (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
>> errmsg("could not import the requested snapshot"),
>> - errdetail("The source transaction %u is not running anymore.",
>> - sourcexid)));
>> + errdetail("The source virtual transaction %d/%u is not running anymore.",
>> + sourcevxid->backendId, sourcevxid->localTransactionId)));
>
> Hm, this is a harder to read. Wonder if we should add a pid field, that'd
> make it a bit easier to interpret?
>

Agreed, see attached. We have to pass the pid around a bit but I don't
think it's too bad.

One thing I don't like is the GetLastLocalTransactionId() that I had to
add because we clear the proc->lxid before we get to AtEOXact_Snapshot()
but I don't have better solution there.

--
Petr Jelinek http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services