mod_lsapi issues

[david]

mod_lsapi

[david]

Installed and excited for this handler. Some websites work great, others I am receiving a flood of 500 errors. error_log shows a flood of the following:

Error sending request (lsphp is killed?): ReceiveLSHeader: nothing to read from backend socket
Child process with pid: xxxx was killed by signal: 11, core dump: 0

[bogdan.sh]

Most common reason for signal 11 is mix of caching engines (APC with Zend etc.) however it should act same way with mod_lsapi and without it.

I worry it will not be easy to track down signal 11, it requires some deep debugging tools like forcing core dump creation, installing php-devel* packages and debugging with gdb .

[david]

Thank you for your quick response Bogdan!

This is with an install with xcache and Zend Guard via easyapache. Do you have a different recommendation?

Thank you!

[david]

A little bit different one on another server:

[Mon Dec 15 21:37:06 2014] [error] [client xxx.xxx.xxx.xxx] [host http://www.somedomaincom] Error sending request(GET / HTTP/1.1); uri(/) content-length(0) (lsphp is killed?): ReceiveLSHeader: nothing to read from backend socket

[david]

I cannot possibly be the only person encountering these errors.

[david]

Did you solve this?
Ive seen this with opcache on PHP 5.2 with ioncube/zend encoded websites - you should turn off the opcache modules for those.
The compiling options don include PECL extensions you might have enabled, so "su - <username> and then run php -m" please.

[m.rahnev]

The same error occurred.

[Mon Jul 31 13:16:57.865922 2017] [lsapi:error] [pid 482112:tid 140427902523136] [client XXX.XXX.XXX.XXX:55769] [host [URL]http://www.example.com[/URL]] Backend error on sending request(GET /contact HTTP/1.1); uri(/contact) content-length(0) (lsphp is killed?): ReceiveAckHdr: nothing to read from backend socket, referer: [URL]http://www.example.com/index[/URL]

(480239): Child process with pid: 485461 was killed by signal: 15, core dump: 0

I activated core dumps in lsphp and following appeared in apache error_log:

lsphp master: worker is killed by SEGV signal - increase PMEM limit

Set limits to unlimited and still crashing. Opcache is also disabled. Any ideas how to solve it?

[bogdan.sh]

There is no easy way to help here. The coredump should be investigated with gdb, perfectly - by our developers.

Better to create support ticket with us.

[morten]

Did you find the cause of this?
On our LiteSpeed server I see in stderr.log alot of these all the time:
2017-11-20 15:47:06.058 [STDERR] (98314): Child process with pid: 113954 was killed by signal: 11, core dump: 0
lsphp master: worker is killed by SEGV signal - increase PMEM limit
2017-11-20 15:47:07.065 [STDERR] (113958): Child process with pid: 113959 was killed by signal: 11, core dump: 0
lsphp master: worker is killed by SEGV signal - increase PMEM limit
2017-11-20 15:47:11.060 [STDERR] (51769): Child process with pid: 114313 was killed by signal: 11, core dump: 0
lsphp master: worker is killed by SEGV signal - increase PMEM limit
2017-11-20 15:47:15.031 [STDERR] (51270): Child process with pid: 114391 was killed by signal: 11, core dump: 0
lsphp master: worker is killed by SEGV signal - increase PMEM limit
2017-11-20 15:47:15.033 [STDERR] (51211): Child process with pid: 114386 was killed by signal: 11, core dump: 0
lsphp master: worker is killed by SEGV signal - increase PMEM limit

How can I easily find which account PMEM limit should be increased?
I cannot seem to find any high users in LVE either.

[ighertesco]

Hello Morten,

If I understood you correctly, you are using LiteSpeed? The reasons could be different on LiteSpeed and Apache with mod_lsapi. It is better to submit a ticket to our support team: https://cloudlinux.zendesk.com/hc/en-us/requests/new