Analyze ALOHA’s HAProxy logs with halog

Being able to analyze logs generated by the ALOHA Load-Balancer stored in a third party syslog server.


Objective

Being able to analyze logs generated by the ALOHA Load-Balancer stored in a third party syslog server.


Complexity

4


Versions

v4.2 and later

ALOHA load balancer How to analyze HAProxy logs with HAlog tool


Synopsis

The ALOHA Load-Balancer, when used in layer 7 mode, generates very verbose log lines. It can store them in a circular buffer in memory. Unfortunately, the memory allocated to store the logs is limited and can only be used for instant troubleshooting.

To analyze an issue which occured a few minutes ago, or to get tendances on server or application response time, it is not enough.

To analyze logs, HAProxy comes with a small tool called halog that we can install and use on the server which stores the logs generated by the ALOHA.


ALOHA remote syslog configuration

In the WUI do the following, go in the Services tab, then click on the Edit icon from the syslog traffic line

Add a line server in the configuration, like in the example below:

server 192.168.10.26:514

Then click on the restart icon .

The ALOHA Load-Balancer emits two types of logs:

  • traffic log: connection log, very verbose.
    They are sent with local0 facility and info severity.
  • event log: events occurring on frontends, backends, servers, etc...
    They are sent with local0 facility and notice severity.

We may want to log them differently in the file system because each type of log can be used for different purpose.


Linux syslog server configuration

syslog-ng

Syslog-ng is one of the most powerful syslog server.

For syslog-ng, you have to define a source, a filter and a destination, like in the example below:

# tells syslog - ng to listen on its external IP
source s_net { udp ( ip ("192.168.10.26") port (514)); };

# where to write the logs
# traffic logs
destination d_aloha_traffic { file ("/ var / log / aloha / traffic . log "
	create_dirs ( yes )); };
# event logs
destination d_aloha_events { file ("/ var / log / aloha / events . log "
	create_dirs ( yes )); };

# ALOHA traffic logs are emitted with facilty local0 and level info
filter f_aloha_traffic { facility ( local0 ) and level ( info ); };
# ALOHA event logs are emitted with facilty local0 and level notice
filter f_aloha_events { facility ( local0 ) and level ( notice ); };

# traffic logging
log { source ( s_net ); filter ( f_aloha_traffic );
	destination ( d_aloha_traffic ); };
# events logging
log { source ( s_net ); filter ( f_aloha_events );
	destination ( d_aloha_events ); };

rsyslog

rsyslog is one of the most used syslog server, since it’s installed by default on the main Linux distribution.

For rsyslog, we have to enable the network socket and route ALOHA log lines too.

In order to make rsyslog listen on the network, uncomment the two line below in the file /etc/rsyslog.conf:

$ModLoad imudp
$UDPServerRun 514

In order to route syslog messages to different files, add the two line below to the end of the /etc/rsyslog.conf file:

# ALOHA logs traffic with facility local0 and severity info
local0 . info -/ var / log / aloha / traffic . log
# ALOHA logs events with facility local0 and severity notice
local0 . notice -/ var / log / aloha / events . log

logrotate

When adding new log files, it’s a good idea to rotate them as well as to delete oldest files. This is the role of logrotate.

Create a new file called aloha in logrotate’s configuration directory /etc/logrotate.d:

/ var / log / aloha /*. log
{
	rotate 31
	daily
	missingok
	notifempty
	delaycompress
	compress
	sharedscripts
	postrotate
		invoke - rc . d rsyslog reload > / dev / null
	endscript
}
In the example above, replace rsyslog by syslog-ng, depending which syslog server you’re running

HALog installation

HALog is a small and very powerful tool to analyze ALOHA’s log lines.

Installation is pretty simple, as described bellow:

cd / usr / src
wget http :// haproxy .1 wt . eu / download /1.5/ src / devel / haproxy -1.5 - dev11 . tar . gz
tar xzf haproxy -1.5 - dev11 . tar . gz
cd haproxy -1.5 - dev11 / contrib / halog
make
cp halog / usr / bin /

Analyzing ALOHA’s logs

Now we have ALOHA’s log and halog in the same server we can run some analyze on them.

List servers by number of requests treated

The command below lists the servers by the number of requests they treated. The server name is prefixed by the backend name.

The eighth columns "tot_req" gives the number of requests treated by the server.

cat traffic . log | halog - srv -H -q | awk ’ NR ==1; NR > 1 { print $0 | " sort -n -r -k 9"} ’ | column -t
# srv_name 1 xx 2 xx 3 xx 4 xx 5 xx other tot_req req_ok pct_ok avg_ct avg_rt
dynamic / server1 0 3510 0 7 0 0 3517 3517 100.0 1495 1747
dynamic / server2 0 3516 0 0 0 0 3516 3516 100.0 1372 1776

List servers by response time

The command below lists the servers by response time. The server name is prefixed by the backend name.

The response time is in milliseconds and the latest columns "avg_rt" gives the average response time for all the URLs forwarded to this server in this backend.

cat traffic . log | halog - srv -H -q | awk ’ NR ==1; NR > 1 { print $0 | " sort -n -r -k 12"} ’ | column -t
# srv_name 1 xx 2 xx 3 xx 4 xx 5 xx other tot_req req_ok pct_ok avg_ct avg_rt
dynamic / server2 0 3516 0 0 0 0 3516 3516 100.0 1372 1776
dynamic / server1 0 3510 0 7 0 0 3517 3517 100.0 1495 1747
It is a best practice to split dynamic and static traffic: you would see the server response time for each type of traffic

List servers by application errors: HTTP status code 5xx

The command below lists the servers by number of application errors. The server name is prefixed by the backend name.

The sixth column "5xx" gives the number of application errors generated by the server.

cat traffic . log | halog - srv -H -q | awk ’ NR ==1; NR > 1 { print $0 | " sort -n -r -k 6"} ’ | column -t
# srv_name 1 xx 2 xx 3 xx 4 xx 5 xx other tot_req req_ok pct_ok avg_ct avg_rt
dynamic / server2 0 3516 0 0 0 0 3516 3516 100.0 1372 1776
dynamic / server1 0 3510 0 7 0 0 3517 3517 100.0 1495 1747
It is a best practice to split your applications per backend, that way you will see whose application generates errors on which server

List servers by errors

The command below lists the servers by number of errors not related to the application. The server name is prefixed by the backend name.

cat traffic . log | halog - srv -H -q | awk ’ NR ==1; NR > 1 { print $0 | " sort -n -r -k 5"} ’ | column -t
# srv_name 1 xx 2 xx 3 xx 4 xx 5 xx other tot_req req_ok pct_ok avg_ct avg_rt
dynamic / server1 0 3510 0 7 0 0 3517 3517 100.0 1495 1747
dynamic / server2 0 3516 0 0 0 0 3516 3516 100.0 1372 1776

List URLs by server computation time

The command below lists the URLs by the average computation time, whatever the server which treated it.

The sixth column "okavg" provides the URL average computation time in milliseconds.

cat traffic . log | halog - ut -H -q | column -t
# req err ttot tavg oktot okavg url
1004 0 6609819 6583 6609819 6583 /3 s . php
2006 0 2771766 1381 2771766 1381 / health . php
2008 0 1601026 797 1601026 797 / fast . php
1004 0 1003335 999 1003335 999 / mega . php
1004 0 406830 405 406830 405 / health . html
7 0 19 2 19 2 / favicon . ico

List URLs by errors

The command below lists the URLs by the number of errors they have generated, whatever the server which treated it or the type of error.

The second column "err" provides the number of errors generated by the given URL (latest column).

cat traffic . log | halog - ue -H -q | column -t
# req err ttot tavg oktot okavg url
1004 0 1003335 999 1003335 999 / mega . php
2006 0 2771766 1381 2771766 1381 / health . php
1004 0 406830 405 406830 405 / health . html
7 0 19 2 19 2 / favicon . ico
2008 0 1601026 797 1601026 797 / fast . php
1004 0 6609819 6583 6609819 6583 /3 s . php

List URLs by missing files: HTTP status code 404

The command below lists the URLs by the number of missing files error they have generated, whatever the server which treated it.

The first column "req" provides the number of 404 returned for the given URL (latest column).

cat traffic . log | halog -u -H -q - hs 404 | column -t
# req err ttot tavg oktot okavg url
7 0 19 2 19 2 / favicon . ico

List URLs by number of request

The command below lists the URLs by the number of time they have been requested on the platform.

The first column "req" provides the number of time the URLs was called.

cat aloha . log | halog -u -H -q | awk ’ NR ==1; NR > 1 { print $0 | " sort -n -r -k 1"} ’ | column -t
# req err ttot tavg oktot okavg url
2008 0 1601026 797 1601026 797 / fast . php
2006 0 2771766 1381 2771766 1381 / health . php
1004 0 6609819 6583 6609819 6583 /3 s . php
1004 0 406830 405 406830 405 / health . html
1004 0 1003335 999 1003335 999 / mega . php
7 0 19 2 19 2 / favicon . ico