Zabbix Documentation 5.4

Zabbix can be used for centralized monitoring and analysis of log files with/without log rotation support.

Notifications can be used to warn users when a log file contains certain strings or string patterns.

To monitor a log file you must have:


Verify agent parameters
Item configuration

Configure a log monitoring item.

All mandatory input fields are marked with a red asterisk.

Specifically for log monitoring items you enter:

Type Select Zabbix agent (active) here.
Key Use one of the following item keys:
log[] or logrt[]:
These two item keys allow to monitor logs and filter log entries by the content regexp, if present.
For example: log[/var/log/syslog,error] . Make sure that the file has read permissions for the 'zabbix' user otherwise the item status will be set to 'unsupported'.
log.count[] or logrt.count[]:
These two item keys allow to return the number of matching lines only.
See supported Zabbix agent item key section for details on using these item keys and their parameters.
Type of information Select:
For log[] or logrt[] items — Log ;
For log.count[] or logrt.count[] items — Numeric (unsigned) .
If optionally using the output parameter, you may select the appropriate type of information other than Log .
Note that choosing a non-Log type of information will lead to the loss of local timestamp.
Update interval (in sec) The parameter defines how often Zabbix agent will check for any changes in the log file. Setting it to 1 second will make sure that you get new records as soon as possible.
Log time format In this field you may optionally specify the pattern for parsing the log line timestamp.
If left blank the timestamp will not be parsed.
Supported placeholders:
* y: Year (0001-9999)
* M: Month (01-12)
* d: Day (01-31)
* h: Hour (00-23)
* m: Minute (00-59)
* s: Second (00-59)
For example, consider the following line from the Zabbix agent log file:
“ 23480:20100328:154718.045 Zabbix agent started. Zabbix 1.8.2 (revision 11211).”
It begins with six character positions for PID, followed by date, time, and the rest of the line.
Log time format for this line would be “pppppp:yyyyMMdd:hhmmss”.
Note that “p” and “:” chars are just placeholders and can be anything but “yMdhms”.

Important notes

Extracting matching part of regular expression

Sometimes we may want to extract only the interesting value from a target file instead of returning the whole line when a regular expression match is found.

Since Zabbix 2.2.0, log items have the ability to extract desired values from matched lines. This is accomplished by the additional output parameter in log and logrt items.

Using the 'output' parameter allows to indicate the “capturing group” of the match that we may be interested in.

should allow returning the entry count as found in the content of:

Only the number will be returned because \1 refers to the first and only capturing group: ([0-9]+).

And, with the ability to extract and return a number, the value can be used to define triggers.

Using maxdelay parameter

The 'maxdelay' parameter in log items allows ignoring some older lines from log files in order to get the most recent lines analyzed within the 'maxdelay' seconds.

By default items for log monitoring follow all new lines appearing in the log files. However, there are applications which in some situations start writing an enormous number of messages in their log files. For example, if a database or a DNS server is unavailable, such applications flood log files with thousands of nearly identical error messages until normal operation is restored. By default, all those messages will be dutifully analyzed and matching lines sent to server as configured in log and logrt items.

Built-in protection against overload consists of a configurable 'maxlines' parameter (protects server from too many incoming matching log lines) and a 4*'maxlines' limit (protects host CPU and I/O from overloading by agent in one check). Still, there are 2 problems with the built-in protection. First, a large number of potentially not-so-informative messages are reported to server and consume space in the database. Second, due to the limited number of lines analyzed per second the agent may lag behind the newest log records for hours. Quite likely, you might prefer to be sooner informed about the current situation in the log files instead of crawling through old records for hours.

The solution to both problems is using the 'maxdelay' parameter. If 'maxdelay' 0 is specified, during each check the number of processed bytes, the number of remaining bytes and processing time is measured. From these numbers the agent calculates an estimated delay — how many seconds it would take to analyze all remaining records in a log file.

If the delay does not exceed 'maxdelay' then the agent proceeds with analyzing the log file as usual.

If the delay is greater than 'maxdelay' then the agent ignores a chunk of a log file by “jumping” over it to a new estimated position so that the remaining lines could be analyzed within 'maxdelay' seconds.

Note that agent does not even read ignored lines into buffer, but calculates an approximate position to jump to in a file.

The fact of skipping log file lines is logged in the agent log file like this:

The “to byte” number is approximate because after the “jump” the agent adjusts the position in the file to the beginning of a log line which may be further in the file or earlier.

Depending on how the speed of growing compares with the speed of analyzing the log file you may see no “jumps”, rare or often “jumps”, large or small “jumps”, or even a small “jump” in every check. Fluctuations in the system load and network latency also affect the calculation of delay and hence, “jumping” ahead to keep up with the “maxdelay” parameter.

Setting 'maxdelay' 'update interval' is not recommended (it may result in frequent small “jumps”).

Notes on handling 'copytruncate' log file rotation

logrt with the copytruncate option assumes that different log files have different records (at least their timestamps are different), therefore MD5 sums of initial blocks (up to the first 512 bytes) will be different. Two files with the same MD5 sums of initial blocks means that one of them is the original, another — a copy.

logrt with the copytruncate option makes effort to correctly process log file copies without reporting duplicates. However, things like producing multiple log file copies with the same timestamp, log file rotation more often than logrt[] item update interval, frequent restarting of agent are not recommended. The agent tries to handle all these situations reasonably well, but good results cannot be guaranteed in all circumstances.

Actions if communication fails between agent and server

Each matching line from log[] and logrt[] item and a result of each log.count[] and logrt.count[] item check requires a free slot in the designated 50% area in the agent send buffer. The buffer elements are regularly sent to server (or proxy) and the buffer slots are free again.

While there are free slots in the designated log area in the agent send buffer and communication fails between agent and server (or proxy) the log monitoring results are accumulated in the send buffer. This helps to mitigate short communication failures.

During longer communication failures all log slots get occupied and the following actions are taken: