PHP mail causing PHP5-FPM to time out

winston

It seems that every time PHP calls mail() it will make the page load forever, and eventually leading to a 504 NGINX timeout.

I have little to no clue of where to begin debugging this, help would be very appreciated.

Kairus

Anything in your php log files?

winston

I think it has something to do with the sendmail package (this is Ubuntu 12.04 x86 btw)

service sendmail restart

It takes about 2 minutes for the following to say [OK]

• Restarting Mail Transport Agent (MTA) sendmail

winston

cat /var/log/php5-fpm.log shows nothing besides restarts

Wintereise

You're checking the wrong log.

Anyway, fpm echos errors to stdout, so nginx catches them on the fcgi interface.

Look at your error log for that specific vhost in nginx.

winston

2012/09/30 01:50:34 [error] 27470#0: *87576 upstream timed out (110: Connection timed out) while reading response header from upstream, client: myip, server: mydomain.yay, request: "POST /?action=login HTTP/1.0", upstream: "fastcgi://127.0.0.1:9000", host: "mydomain.yay", referrer: "http://mydomain.yay/"

Wintereise

Nothing specific about mailing?

Could it be that it's simply being overloaded? (Though, I guess that'd notify you in php5-fpm.log)

I think you can change the verbosity of the log in php5-fpm.conf, try doing that.

jhadley

This isn't WHMCS by any chance?

winston

@Wintereise No way it's overloaded, only happens when mail is sent. SMTP mail works fine. service sendmail restart also hangs for several minutes. Also, can't find php5-fpm.conf..

locate php5-fpm.conf

/var/lib/dpkg/info/php5-fpm.conffiles

winston

@jhadley said: This isn't WHMCS by any chance?

Nope :(

winston

May it have something to do with changing the VPS hostname?

qjq

the configuration is in the pool directory probably something like /etc/php/fpm/pool.d/www.conf

Wintereise

; Log level
; Possible Values: alert, error, warning, notice, debug
; Default Value: notice
;log_level = notice

This is the directive I was talking about. And no, you don't have this in the pool config, but rather the global php-fpm configuration file.

For me, it's

root@gecos:/mnt# locate php-fpm.conf
/etc/php5/fpm/php-fpm.conf

there.

likexian

Try using mail() in cli. for example, php -r "mail(...);" and try to find out it works or not.

MikHo

Does sendmail actually start ?

If its not running that would explain the timeout for php and we can shift focus to troubleshooting sendmail. Instead.

Wintereise

You can try out other mail transfer agents too, if the problem is indeed with Sendmail.

winston

@Wintereise it was actually where @qjq said it was, should I be setting it to error or debug?

@likexian it hangs just as long

@MikHo yeah, it just takes a few minutes for it to start up

@Wintereise anything you would recommend? Something that would require little setting up on my end, as all I wish to do is send mail without pages dying :<

sleddog

Does the maillog show any warnings/errors, especially when you restart sendmail?

I haven't used sendmail for years, but if the old hostname is set in the sendmail config, and that hostname no longer resolves, then there will likely be issues with sendmail....

winston

Sep 30 16:00:01 saturn sm-msp-queue[12177]: My unqualified host name (saturn) unknown; sleeping for retry Sep 30 16:01:01 saturn sm-msp-queue[12177]: unable to qualify my own domain name (saturn) -- using short name

winston

@sleddog ding ding ding! Updated /etc/hosts and it worked, thanks!

Wintereise

Lol, I believe the sendmail installer (when run via apt) actually tells you if your server hostname resolves and if it should be updated.

Glad to see you've gotten it fixed though :)

sleddog

@winston said: ding ding ding! Updated /etc/hosts and it worked, thanks!

Cool you fixed it :)