Re: [exim] Failed to get write lock

On 2009-02-18 at 02:25 +0800, W B Hacker wrote:
> Failure beyond ten minutes that retry will *ever* recover from is very
> much the exception.

Rare != Should-Not-Be-Accounted-For

Eg, system crashes, is configured to fsck before resuming mail
operations on the basis that mail load means background fsck will never
catch up, takes 20 or 30 minutes to come back to service.

If it's a corporate server, it should be part of a farm and this won't
matter, but private systems or small non-profits running on a
shoe-string budget?

System has hardware failure, needs someone to drive out to premises,
replace hardware, kick the system back to life; outage of, say, 4 hours.
If you're handling interactive mail, it might be reasonable to bounce
for the duration, but if you're running a mailing-list it's probably a
bad idea to unsubscribe or cancel delivery for the addresses on that
remote server.

I'd suggest that at the very least you keep the retries longer for mails
coming from a mailing-list manager. To do this, note that the pattern
in the retry rules is an expanded string, so can be made conditional
upon Exim variables of your choice.

All this ignorable if you only care about mail delivery to people
running redundant systems and are willing to assume that all your
business partners run things so well. Most of us don't assume that
much and support communication with others.

-Phil