On Thu, Nov 20, 2008 at 1:34 PM, Dean Brooks <dean@???> wrote:
> On Thu, Nov 20, 2008 at 10:34:55AM -0800, Brent Jones wrote:
>> I have a small load server (70k deliveries a day) and thought I'd try
>> using queue_only, with queue_run_max = 15 to see if it makes things a
>> bit more efficient.
>
> Couple of things:
>
> 1. Exim only starts a queue runner every nn minutes, where nn is
> specified on the "-q" flag on the startup command line of Exim daemon
> itself. This is commonly set to "-q15m", but you can certainly lower
> it to something lower, such as "-q1m". Just watch the system load
> carefully. This is typically set in your /etc/rc2.d startup directory
> where Exim daemon is first started.
>
I have it set to -q1m
> 2. When the "-q" time limit above is reached, it only spawns a single
> extra queue runner. Therefore, if you set your daemon to "-q1m",
> it could take 15 minutes from restart before all 15 queue runners
> are running (per your queue_run_max=15 setting).
>
That I didn't know, I thought it would spawn 15 runners, and each
would process the queue and be off.
> 3. Remember that queued messages will be skipped if they are pending
> a retry due to a previous temporary failure. As a result, sometimes a
> queue runner will start and skip over a large number of messages.
> That is normal if those messages haven't reached their retry time yet.
>
Yah, anything that ends up in the retry queue will eventually get
frozen and timeout. Lots of messages get skipped.
> 4. After a day or so running, you'll likely not see all 15 queue
> runners running at same time unless you have a lot of undeliverables
> on your queue. The reason is that most messages will not have hit
> their retry time yet, and the queue runner will fly through all the
> messages and then end. Once a queue runner has run through all the
> messages, it quietly ends.
>
> 5. Setting queue_only may not help performance much if your box
> is already sitting idle most of the time. The most it will do is
> delay processing of your message. If your machine isn't heavily loaded
> and you don't fear spikes of heavy use, its safe to revert back
> to immediate processing of incoming messages. Setting queue_only
> does keep the load a bit more consistent though.
>
> Hope this helps some.
>
Helps a lot, thank you for explaining those options.
I was hoping to make sure the load was steady even if we got a big
spike, but it like wont be an issue for a while.
It was just odd seeing deliveries not be captured by the queue runners
and delivered the way I thought it would work, seeing some normal
deliveries sit in the queue even with multiple queue runners, passing
up those messages even if they hadn't been tried yet.
I'll revert from queue_only, and just let things operate as they were
which worked perfectly.
Thanks
> --
> Dean Brooks
> dean@???
>
> --
> ## List details at http://lists.exim.org/mailman/listinfo/exim-users
> ## Exim details at http://www.exim.org/
> ## Please use the Wiki with this list - http://wiki.exim.org/
>
--
Brent Jones
brent@???
