[ZBX-23356] 100% cache-hit, loosing events Created: 2023 Sep 04  Updated: 2023 Sep 05  Resolved: 2023 Sep 05

Status: Closed
Project: ZABBIX BUGS AND ISSUES
Component/s: None
Affects Version/s: 6.0.20
Fix Version/s: None

Type: Problem report Priority: Trivial
Reporter: Kostas Zourkas Assignee: Zabbix Support Team
Resolution: Commercial support required Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified
Environment:

We have a single zabbix server installation, with two proxies. The server is running on a VM (vmware) using a PGSQL with timescaleDB on a different system. The db is around 650G. The system is managing 2700 hosts, 144k items, 81k triggers.


Attachments: PNG File Internal_process.png     PNG File cache_hit.png     PNG File cpu.png     PNG File data_gathering.png     PNG File data_handling.png     PNG File poller.png     PNG File queue_depth.png     PNG File utilization.png    

 Description   

The system is low on CPU and has more that 50% availiable memory. 

The problem is that when we start the server, it takes about 3 hours to reach 100% History write cache and after that the poller utilization goes to 100%. 

After that the system is not updating any graphs. 

The PGSQL db has low CPU and mem usage, we did not found anything out of the ordinary. The disk file system sits on an NMVE enterprise disk array.

When we stop the zabbix server it takes roughly 6 hours to stop the process as it needs to write data to the DB. 

We need to find out what is wrng with the setup.



 Comments   
Comment by Edgar Akhmetshin [ 2023 Sep 05 ]

The problem is that when we start the server, it takes about 3 hours to reach 100% History write cache and after that the poller utilization goes to 100%.
After that the system is not updating any graphs.

History Write Cache is a buffer between database and Zabbix Server write workflow, 100% utilisation of the write cache means that database is not able to keep up with workload. Ask DBA to optimise database performance.

Please be advised that this section of the tracker is for bug reports only. The case you have submitted can not be qualified as one, so please reach out to [email protected] for commercial support (https://zabbix.com/support) or consultancy services. Alternatively, you can also use our IRC channel or community forum (https://www.zabbix.com/forum) for assistance. With that said, we are closing this ticket. Thank you for understanding.

Generated at Sat Apr 12 11:01:49 EEST 2025 using Jira 9.12.4#9120004-sha1:625303b708afdb767e17cb2838290c41888e9ff0.