Re: Zombies roaming the server

On Feb 18, 2008 10:42 PM, Yoshiyuki Asaba 
<y-asaba-r5vX20e9KLfqq2nvvmkE/A@xxxxxxxxxxxxxxxx> wrote:
> > Now, instead of dropping the secondary backend, it just stops
> > receiving connections and its children turn into zombies.
>
> Thank you for the report. I've fixed it.
> Could you try CVS HEAD version?

Thank you, I'm recompiling and will stop and resync the db as soon as I can.
However, after I took the 'RESET ALL' from the query list, it did not
happened again. I might be lucky.
I noticed some error messages that brought the db offline for small
periods. There are lots of them, and after some time, it seems the
problem solved itself. Here are some of them:

Looks like the frontend crashed for some reason:

2008-02-18 12:14:50 ERROR: pid 10969: ProcessFrontendResponse: failed
to read kind from frontend. fronend abnormally exited
2008-02-18 12:14:50 LOG:   pid 10969: do_child: exits with status 1 due to error


There are lots of these between 04:14 and 05:12. Really lots.

2008-02-19 04:14:06 ERROR: pid 8547: pool_check_fd: data is not ready
tp->tv_sec 5 tp->tp_usec 5000000
2008-02-19 04:14:06 ERROR: pid 8547: pool_read: pool_check_fd failed (Success)
2008-02-19 04:14:06 ERROR: pid 8547: pool_process_query: failed to
read kind from 0 th backend
2008-02-19 04:14:06 LOG:   pid 8547: do_child: exits with status 1 due to error


After that, the database goes offline with messages like this (there
are a few of the previous too):

2008-02-19 05:12:44 ERROR: pid 10141: pool_read_kind: kind does not
match between master(69) slot[1] (83)
2008-02-19 05:12:44 ERROR: pid 10141: pool_do_auth: failed to read
kind before BackendKeyData


And then, after 05:25, all goes back to normal.

Is there anything I can do to help you debug?

-- 
Bruno Lustosa <bruno-God0kL5/+C3R7s880joybQ@xxxxxxxxxxxxxxxx>
http://www.lustosa.net/