Posts Tagged ‘DEBUG’

Debug and fix Virtuozzo / KVM broken Hypervisor error: ‘PrlSDKError(‘SDK error: 0x80000249: Unable to connect to Virtuozzo. You may experience a connection problem or the server may be down.’ on CentOS Linux howto

Thursday, January 28th, 2021

Reading Time: 6minutes

fix-sdkerror-virtuozzo-kvm-how-to-debug-problems-with-hypervisor-host-linux

I've recently yum upgraded a CentOS Linux server runinng Virtuozzo kernel and Virtuozzo virtualization Virtual Machines to the latest available CentOS Linux release 7.9.2009 (Core) just to find out after the upgrade there was issues with both virtuozzo (VZ) way to list installed VZ enabled VMs reporting Unable to connect to Virtuozzo error like below:
 

[root@CENTOS etc]# prlctl list -a
Unable to connect to Virtuozzo. You may experience a connection problem or the server may be down. Contact your Virtuozzo administrator for assistance.


Even the native QEMU KVM VMs installed on the Hypervisor system failed to work to list and bring up the VMs producing another unexplainable error with virsh unable to connect to the hypervisor socket

[root@CENTOS etc]# virsh list –all
error: failed to connect to the hypervisor
error: Failed to connect socket to '/var/run/libvirt/libvirt-sock': No such file or directory


In dmesg cmd kernel log messages the error found looked as so:

[root@CENTOS etc]# dmesg|grep -i sdk


[    5.314601] PrlSDKError('SDK error: 0x80000249: Unable to connect to Virtuozzo. You may experience a connection problem or the server may be down. Contact your Virtuozzo administrator for assistance.',)

To fix it I had to experiment a bit based on some suggestions from Google results as usual and what turned to be the cause is a now obsolete setting for disk probing that is breaking libvirtd

Disable allow_disk_format_probing in /etc/libvirt/qemu.conf

The fix to PrlSDKError('SDK error: 0x80000249: Unable to connect to Virtuozzo comes to commenting a parameter inside 

/etc/libvirt/qemu.conf

which for historical reasons seems to be turned on by default it is like this

allow_disk_format_probing = 1


Resolution is to either change the value to 0 or completely comment the line:

[root@CENTOS etc]# grep allow_disk_format_probing /etc/libvirt/qemu.conf
# If allow_disk_format_probing is enabled, libvirt will probe disk
#allow_disk_format_probing = 1
#allow_disk_format_probing = 1


Debug problems with Virtuozzo services and validate virtualization setup

What really helped to debug the issue was to check the extended status info of libvirtd.service vzevent vz.service libvirtguestd.service prl-disp systemd services

[root@CENTOS etc]# systemctl -l status libvirtd.service vzevent vz.service libvirtguestd.service prl-disp

Here I had to analyze the errors and googled a little bit about it


Once this is changed I had to of course restart libvirtd.service and rest of virtuozzo / kvm services

[root@CENTOS etc]# systemctl restart libvirtd.service ibvirtd.service vzevent vz.service libvirtguest.service prl-disp


Another useful tool part of a standard VZ install that I've used to make sure each of the Host OS Hypervisor components is running smoothly is virt-host-validate(tool is part of libvirt-client rpm package)

[root@CENTOS etc]# virt-host-validate
  QEMU: Checking for hardware virtualization                                 : PASS
  QEMU: Checking if device /dev/kvm exists                                   : PASS
  QEMU: Checking if device /dev/kvm is accessible                            : PASS
  QEMU: Checking if device /dev/vhost-net exists                             : PASS
  QEMU: Checking if device /dev/net/tun exists                               : PASS
  QEMU: Checking for cgroup 'memory' controller support                      : PASS
  QEMU: Checking for cgroup 'memory' controller mount-point                  : PASS
  QEMU: Checking for cgroup 'cpu' controller support                         : PASS
  QEMU: Checking for cgroup 'cpu' controller mount-point                     : PASS
  QEMU: Checking for cgroup 'cpuacct' controller support                     : PASS
  QEMU: Checking for cgroup 'cpuacct' controller mount-point                 : PASS
  QEMU: Checking for cgroup 'cpuset' controller support                      : PASS
  QEMU: Checking for cgroup 'cpuset' controller mount-point                  : PASS
  QEMU: Checking for cgroup 'devices' controller support                     : PASS
  QEMU: Checking for cgroup 'devices' controller mount-point                 : PASS
  QEMU: Checking for cgroup 'blkio' controller support                       : PASS
  QEMU: Checking for cgroup 'blkio' controller mount-point                   : PASS
  QEMU: Checking for device assignment IOMMU support                         : PASS
  QEMU: Checking if IOMMU is enabled by kernel                               : WARN (IOMMU appears to be disabled in kernel. Add intel_iommu=on to kernel cmdline arguments)
   LXC: Checking for Linux >= 2.6.26                                         : PASS
   LXC: Checking for namespace ipc                                           : PASS
   LXC: Checking for namespace mnt                                           : PASS
   LXC: Checking for namespace pid                                           : PASS
   LXC: Checking for namespace uts                                           : PASS
   LXC: Checking for namespace net                                           : PASS
   LXC: Checking for namespace user                                          : PASS
   LXC: Checking for cgroup 'memory' controller support                      : PASS
   LXC: Checking for cgroup 'memory' controller mount-point                  : PASS
   LXC: Checking for cgroup 'cpu' controller support                         : PASS
   LXC: Checking for cgroup 'cpu' controller mount-point                     : PASS
   LXC: Checking for cgroup 'cpuacct' controller support                     : PASS
   LXC: Checking for cgroup 'cpuacct' controller mount-point                 : PASS
   LXC: Checking for cgroup 'cpuset' controller support                      : PASS
   LXC: Checking for cgroup 'cpuset' controller mount-point                  : PASS
   LXC: Checking for cgroup 'devices' controller support                     : PASS
   LXC: Checking for cgroup 'devices' controller mount-point                 : PASS
   LXC: Checking for cgroup 'blkio' controller support                       : PASS
   LXC: Checking for cgroup 'blkio' controller mount-point                   : PASS
   LXC: Checking if device /sys/fs/fuse/connections exists                   : PASS


One thing to note here that virt-host-validate helped me to realize the  fuse (File system in userspace) module kernel support enabled on the HV was missing so I've enabled temporary for this boot with modprobe and permanently via a configuration like so:

# to load it one time
[root@CENTOS etc]#  modprobe fuse
# to load fuse permnanently on next boot

[root@CENTOS etc]#  echo fuse >> /etc/modules-load.d/fuse.conf

Disable selinux on CentOS HV

Another thing was selinux was enabled on the HV. Selinux is really annoying thing and to be honest I never used it on any server and though its idea is quite good the consequences it creates for daily sysadmin work are terrible so I usually disable it. It could be that a Hypervisor Host OS might work just normal with the selinux enabled but just in case I decided to remove it. This is how

[root@CENTOS etc]#  sestatus
SELinux status:                 enabled
SELinuxfs mount:                /sys/fs/selinux
SELinux root directory:         /etc/selinux
Loaded policy name:             targeted
Current mode:                   enforcing
Mode from config file:          enforcing
Policy MLS status:              enabled
Policy deny_unknown status:     allowed
Max kernel policy version:      31

To temporarily change the SELinux mode from targeted to permissive with the following command:

[root@CENTOS etc]#  setenforce 0

Edit /etc/selinux/config file and set the SELINUX mod to disabled

[root@CENTOS etc]# vim /etc/selinux/config
# This file controls the state of SELinux on the system.
# SELINUX= can take one of these three values:
#       enforcing – SELinux security policy is enforced.
#       permissive – SELinux prints warnings instead of enforcing.
#       disabled – No SELinux policy is loaded.
SELINUX=disabled
# SELINUXTYPE= can take one of these two values:
#       targeted – Targeted processes are protected,
#       mls – Multi Level Security protection.
SELINUXTYPE=targeted

Finally rebooted graceously the machine just in case with the good recommended way to reboot servers with shutdown command instead of /sbin/reboot

[root@CENTOS etc]# shutdown -r now

The advantage of shutdown is that it tries to shutdown each service by sending stop requests but usually this takes some time and even a shutdown request could take longer to proccess as each service such as a WebServer application is being waited to close all its network connections etc. |
However if you want to have a quick reboot and you don't care about any established network connections to third party IPs you can go for the brutal old fashioned /sbin/reboot 🙂

Check is MySQL dropping connections or how to check drop out SQL problems and debug problems with slow queries

Thursday, October 26th, 2017

Reading Time: 2minutes

mysql-dropping-connections-how-to-check-command-drop-out-sql-problems-and-debugging-slow-queries

If you have migrated a website or you recently upgraded MySQL and you face occasional problems with dropped connections to remote Web site or application it is likely that the problem is rooted either in Apache being unable to handle multiple connections, e.g. you will get a lot of 503 errors in log or MySQL dropping connections due to someout or MySQL / PHP bug that occurs after packages upgrade.

So how to check is MySQL dropping connections?

Use below command one liner shell script and check for Aborted conncetions a relatively high number is indication MySQL is dropping connections from PHP script or local / remote application such as SOAP or XSLT connects.
 

# echo 'show global status' | mysql -u root -p |grep -i connect
Enter password:
Aborted_connects    710
Connections    49537
Max_used_connections    76
Ssl_client_connects    0
Ssl_connect_renegotiates    0
Ssl_finished_connects    0
Threads_connected    1

Another indicator for MySQL dropping connections is if you find MySQL to have many slow queries, that would mean the slow queries won't be processed on time and there will be consequential connection drop outs.
 

# echo 'show global status' | mysql -u root -p |grep -i slow
Enter password:
Slow_launch_threads    0
Slow_queries    120237


If you find any you can use some MySQL / MariaDB monitoring script such as (mytop / mtop – for more on monitoring MySQL queries in real time I recommend you check out my previous article Monitoring MySQL queries and debugging performance (slow queries))

Disable php notice logging / stop variable warnings in error.log on Apache / Nginx / Lighttpd

Monday, July 28th, 2014

Reading Time: 2minutes

disable_php_notice_warnings_logging_in-apache-nginx-lighttpd
At one of companies where I administrate few servers, we are in process of optimizing the server performance to stretch out the maximum out of server hardware and save money from unnecessery hardware costs and thus looking for ways to make server performance better.

On couple of web-sites hosted on few of the production servers, administrating, I've noticed dozens of PHP Notice errors, making the error.log quickly grow to Gigabytes and putting useless hard drive I/O overhead. Most of the php notice warnings are caused by unitialized php variables.

I'm aware having an unitialized values is a horrible security hole, however the websites are running fine even though the notice warnings and currently the company doesn't have the necessery programmers resource to further debug and fix all this undefined php vars, thus what happens is monthly a couple of hundreds megabytes of useless same php notice warnings are written in error.log.

That  error.log errors puts an extra hardship for awstats which is later generating server access statistics while generating the 404 errors statistics and thus awstats script has to read and analyze huge files with plenty of records which doesn't have nothing to do with 404 error

We found this PHP Notice warnings logged is one of the things we can optimize had to be disabled.

Here is how this is done:
On the servers running Debian Wheezy stable to disable php notices.

I had to change in /etc/php5/apache2/php.inierror_reporting variable.

Setting was to log everything (including PHP critical errors, warning and notices) like so:
 

vi /etc/php5/apache2/php.ini

error_reporting = E_ALL & ~E_DEPRECATED

to

error_reporting = E_COMPILE_ERROR|E_ERROR|E_CORE_ERROR


On CentOS, RHEL, SuSE based servers, edit instead /etc/php.ini.

This setting makes Apache to only log in error.log critical errors, php core dump (thread) errors and php code compilation (interpretation errors)

To make settings take affect on Debian host Apache webserver:

/etc/init.d/apache2 restart

On CentOS, RHEL Linux, had to restart Apache with:

/etc/init.d/httpd restart

For other servers running Nginx and Lighttpd webservers, after changing php.ini:

service nginx reload
service lighttpd restart

To disable php notices errors only on some websites, where .htaccess enabled, you can use also place in website DocumentRoot.htaccess:
 

php_value error_reporting 2039


Other way to disable via .htaccess is by adding to it code:
 

php_flag display_errors off


Also for hosted websites on some of the servers, where .htaccess is disabled, enabling / disabling php notices can be easily triggered by adding following php code to index.php

define('DEBUG', true);

if(DEBUG == true)
{
    ini_set('display_errors', 'On');
    error_reporting(E_ALL);
}
else
{
    ini_set('display_errors', 'Off');
    error_reporting(0);
}

 

Where does Debian GNU / Linux Apache + PHP stores session files?

Tuesday, November 22nd, 2011

Reading Time: < 1minute
In order to debug some PHP session problems on Debian, I needed to check the count of existing session files.
When PHP is compiled from source usually, by default sessions are stored in /tmp directory, however this is not the case on Debian.

Debian’s PHP session directory is different, there the sessions are stored in the directory:

/var/lib/php5

I’ve discovered the session directory location by reading Debian’s cron shell script, which delete session files on every 30 minutes.

Here is the file content:

debian~# cat /etc/cron.d/php5
# /etc/cron.d/php5: crontab fragment for php5
# This purges session files older than X, where X is defined in seconds
# as the largest value of session.gc_maxlifetime from all your php.ini
# files, or 24 minutes if not defined. See /usr/lib/php5/maxlifetime

# Look for and purge old sessions every 30 minutes
09,39 * * * * root [ -x /usr/lib/php5/maxlifetime ] &&
[ -d /var/lib/php5 ] && find /var/lib/php5/ -type f -cmin +$(/usr/lib/php5/maxlifetime) -delete

To check the amount of existing PHP opened session files:

debian:~# ls -1 /var/lib/php5|wc -l
14049

List and get rid of obsolete program core dump files and completely disable core files on FreeBSD

Tuesday, November 1st, 2011

Reading Time: < 1minute
My FreeBSD router has started running out of space, I looked for ways to clean up some space. So I remembered some programs are generating core files while they crash. Some of these files are really huge and ban be from 1Mb to > 1G.

I used find to first list all my produced core files starting from root directory (/) , like so:

find / -name core -exec du -hsc {} ;
....

Having a list of my core files with the respective core file size and after reviewing, I deleted one by one the cores which were there just taking up space.
It’s a wise idea that core dumps file generation on program crash is completely disabled, however I forgot to disable cores, so I had plenty of the cores – (crash files which are handy for debug purposes and fixing the bug that caused the crash).

Further on I used an /etc/rc.confdumpdev=NO , variable which instructs the kernel to not generate core files on program crash:

freebsd# echo 'dumpdev=NO' >> /etc/rc.conf

Next, to make dumpdev=NO , take affect I rebooted the server:

freebsd# shutdown -r now
...

There is a way to instruct every server running daemon to know about the newly set dumpdev=NO by restarting each of the services with their init scripts individually, but I was too lazy to do that.

How to convert any internet Webpage to PDF from command line on GNU/Linux

Friday, September 30th, 2011

Reading Time: 2minutesLinux webpage html to pdf command line convertor wkhtmltopdf

If you're looking for a command line utility to generate PDF file out of any webpage located online you are looking for Wkhtmltopdf
The conversion of webpages to PDF by the tool is done using Apple's Webkit open source render.
wkhtmltopdf is something very useful for web developers, as some webpages has a requirement to produce dynamically pdfs from a remote website locations.
wkhtmltopdf is shipped with Debian Squeeze 6 and latest Ubuntu Linux versions and still not entered in Fedora and CentOS repositories.

To use wkhtmltopdf on Debian / Ubuntu distros install it via apt;

linux:~# apt-get install wkhtmltodpf
...

Next to convert a webpage of choice use cmd:

linux:~$ wkhtmltopdf pc-freak.net pc-freak.net_website.pdf
Loading page (1/2)
Printing pages (2/2)
Done

If the web page to be snapshotted in long few pages a few pages PDF will be generated by wkhtmltopdf
wkhtmltopdf also supports to create the website snapshot with a specified orientation Landscape / Portrait

-O Portrait options to it, like so:

linux:~$ wkhtmltopdf -O Portrait pc-freak.net pc-freak.net_website.pdf

wkhtmltopdf has many useful options, here are some of them:
 

  • Javascript disabling – Disable support for javascript for a website
  • Grayscale pdf generation – Generates PDf in Grayscale
  • Low quality pdf generation – Useful to shrink the output size of generated pdf size
  • Set PDF page size – (A4, Letter etc.)
  • Add zoom to the generated pdf content
  • Support for password HTTP authentication
  • Support to use the tool over a proxy
  • Generation of Table of Content based on titles (only in static version)
  • Adding of Header and Footers (only in static version)

To generate an A4 page with wkhtmltopdf:

wkhtmltopdf -s A4 pc-freak.net/blog/ pc-freak.net_blog.pdf

wkhtmltopdf looks promising but seems a bit buggy still, here is what happened when I tried to create a pdf without setting an A4 page formatting:

linux:$ wkhtmltopdf pc-freak.net/blog/ pc-freak.net_blog.pdf
Loading page (1/2)
OpenOffice path before fixup is '/usr/lib/openoffice' ] 71%
OpenOffice path is '/usr/lib/openoffice'
OpenOffice path before fixup is '/usr/lib/openoffice'
OpenOffice path is '/usr/lib/openoffice'
** (:12057): DEBUG: NP_Initialize
** (:12057): DEBUG: NP_Initialize succeeded
** (:12057): DEBUG: NP_Initialize
** (:12057): DEBUG: NP_Initialize succeeded
** (:12057): DEBUG: NP_Initialize
** (:12057): DEBUG: NP_Initialize succeeded
** (:12057): DEBUG: NP_Initialize
** (:12057): DEBUG: NP_Initialize succeeded
Printing pages (2/2)
Done
Printing pages (2/2)
Segmentation fault

Debian and Ubuntu version of wkhtmltopdf does not support TOC generation and Adding headers and footers, to support it one has to download and install the static version of wkhtmltopdf
Using the static version of the tool is also the only option for anyone on Fedora or any other RPM based Linux distro.