Page created: 8 Dec 2020 |
Page updated: 12 May 2021
The PingDataMetrics server polls all of the monitored servers at a regular interval to collect data it to import into the database management system (DBMS).
When they are available, the PingDataMetrics server fetches new samples through LDAP. The PingDataMetrics server has one dedicated thread taking sample blocks and converting them to the normalized form stored in the DBMS. The import queue’s size is normally near zero, but under certain conditions it can become large. When the PingDataMetrics server starts, it will queue for import all sample blocks still on disk. Blocks that are older than two hours are discarded.
If a monitored server becomes unavailable for an extended period of time, it will continue to queue blocks of samples locally. When it becomes available again, the PingDataMetrics server collection poll of that server will capture hundreds or thousands of sample blocks. The PingDataMetrics server captures the sample blocks at a much faster speed than it can import them, causing the queue to grow for a period of time. If the PingDataMetrics server is stopped, this problem is compounded because all monitored servers will then have a backlog of sample blocks to be imported.