missing store dir handling

[Uwe Kiewel]

in a case of missing/unwriteable store dir, piller is dropping the mail. From my point of view, piller should crash - or queue mails

I ran into data loss due to NFS unavailability:

Mar 16 19:44:15 mailarchiv piler[1655]: 4000000056e9a98926dfed5c001350b1cf07: cannot open: /var/piler/store/00/56e/cf/07/4000000056e9a98926dfed5c001350b1cf07.m Mar 16 19:44:15 mailarchiv piler[1655]: 4000000056e9a98926dfed5c001350b1cf07: error storing message: 4000000056e9a98926dfed5c001350b1cf07.m

later I checked: [root@mailarchiv ~]# pilerget 4000000056e9a98926dfed5c001350b1cf07 [root@mailarchiv ~]#

that means empty result, mail is not archived

[Janos SUTO]

I strongly disagree with you. Firstly, piler shouldn't crash, ever. A managing a queue is possible, however I've decided not doing so. If piler encounters an issue I think it returns 451, which translates as a temporary error. I believe it's the right way to prevent a data loss, forcing the sending party to store the email in its queue until piler can receive it.

So please double check the smtp status code the piler daemon returned, if it's really 2xx (or 5xx) in such a case. And I'd check the logs of the mail server: it should still have the message in question.

[Janos SUTO]

Any update?

[Uwe Kiewel]

It seems to be working as descibed. Issues are handeled based on smtp codes

Von meinem Huawei-Mobiltelefon gesendet

[eXtremeSHOK]

I've done a setup where we use a central postfix to manage the message queue infront of piler.

Then you get all the benefits and advanced queue management features your could require

[Janos SUTO]

Anyway, it wasn't a bug at all.