First off, let me say that I have always considered Zabbix to be a superlative product. That being said...I am running into an issue lately that has me perplexed...
Using Zabbix version 2.4.2, when using filtering by a host group that has 30 hosts, each with 200-300 items, the response when pulling up latest.php is excruciatingly slow (20-40 seconds).
Our install of Zabbix uses low-level discovery (LLD) to populate SNMP items (which is why there are so many items for a given host). Rather than have to manually create specific templates and entries for each device, LLD allows us to create several discovery items (interfaces, storage, etc.).
Note that I have tried partitioning tables, adding indexes, adding more memory to the database server, etc. Data comes back in a reasonable period of time, and there are few queries coming back as slow (over 1 second) in MySQL. After quite a few days of investigation, the bottleneck appears to be **creating** the page, rather than retrieving the information.
Per the Zabbix change logs, prior to 2.2, the latest.php page pulled down only the data it needed to display the items that are expanded (- next to the host items, instead of the + character) -- resulting in a much faster display time.
In our system, we only have configured 905 hosts, with 27,935 items enabled. The numbers do not seem to come close to the maximums published, so we would expect to see better performance.
Two questions:
1. Is anyone else running a large scale system having the same issue?
2. Is there a way to enable the old behavior as a setting within Zabbix, or via a patch?
Our organization is considering switching from Nagios to Zabbix, but need to have comparable performance. Right now, that is not the case for displaying the latest data / current status of items for a given host group.
Help!
Using Zabbix version 2.4.2, when using filtering by a host group that has 30 hosts, each with 200-300 items, the response when pulling up latest.php is excruciatingly slow (20-40 seconds).
Our install of Zabbix uses low-level discovery (LLD) to populate SNMP items (which is why there are so many items for a given host). Rather than have to manually create specific templates and entries for each device, LLD allows us to create several discovery items (interfaces, storage, etc.).
Note that I have tried partitioning tables, adding indexes, adding more memory to the database server, etc. Data comes back in a reasonable period of time, and there are few queries coming back as slow (over 1 second) in MySQL. After quite a few days of investigation, the bottleneck appears to be **creating** the page, rather than retrieving the information.
Per the Zabbix change logs, prior to 2.2, the latest.php page pulled down only the data it needed to display the items that are expanded (- next to the host items, instead of the + character) -- resulting in a much faster display time.
In our system, we only have configured 905 hosts, with 27,935 items enabled. The numbers do not seem to come close to the maximums published, so we would expect to see better performance.
Two questions:
1. Is anyone else running a large scale system having the same issue?
2. Is there a way to enable the old behavior as a setting within Zabbix, or via a patch?
Our organization is considering switching from Nagios to Zabbix, but need to have comparable performance. Right now, that is not the case for displaying the latest data / current status of items for a given host group.
Help!
Comment