Fullscreen
Loading...
 
Skip to main content
1

Monitoring revamp

Context

The Tiki community manages several Tiki-powered sites. Ref:

All these sites are on various servers, and managed by various people. When one site goes down or has poor performance, it is usually not easy to find the root cause and solve it. The main problem is that we are somewhat blind.


Is it a

  • A server issue?
  • A network issue?
  • A Tiki bug?
  • A Tiki misconfiguration?
  • A spike in traffic?
  • A PHP bug? We struggled for over a year for random server crashes, and in the end, it was this: https://bugs.php.net/bug.php?id=71135Question


Different issues require different types of skill set to resolve.

We need better tools and processes. These will help:

  1. GlitchTip
    • Let's add a trigger on slow pages
  2. The upcoming VirtualminQuestion + Debian 10 infrastructure : https://gitlab.com/wikisuite/virtualmin-installerQuestion
  3. Real User Measurement
  4. A monitoring solution that informs the right people in real time. We have selected Zabbix and NetData and we need a better integration with Tiki, keeping it open-ended to integrate with any monitoring system.

Steps

1.1.1. Set up Zabbix (Fabio) and NetData (Horia)

  • Sending alerts to XMPP room: xmpp:monitoring@conference.wikisuite.chat
  • With generic server monitoring (disk space, CPU, etc.)
  • Is OPcache covered? So we avoid this message in Tiki: "Little memory available. Thrashing likely to occur. The values to increase are apc.shm_size (for APC), xcache.size (for XCache) or opcache.memory_consumption (for OPcache)." If Zabbix doesn't do well, we should add a Tiki-specific alert
  • Use NetData in VirtualminQuestion


1.1.2. Improve Tiki code to provide Tiki-specific alert to Zabbix and NetData (help needed)


Here is some code, which needs a review and a revamp:

1.1.3. Improve Tiki manager code to provide specific alerts to Zabbix and NetData (help needed)

1.1.4. Real User Measurement

  • Real User Measurement
  • tiki-performance-stats.php will provide a list of slowest pages, on which we can focus our energy.

History Source Comments
Show PHP error messages