We have configured PMM2 for our cloudsql instances , however as part of maintenance activities the systems have undergone for a restart, after that we are not able to see our DB Metrics
Does that mean we need to remove our instance entries from inventory and add again after doing a tunneling again, or just do a tunneling again so that the metrics gets populated
In either case we are not able to see the metrics in POSTGRES INSTANCE DASH BOARD, however we are able to see the entires in the nodes/agents/services Tab
QAN is working fine, but not able to view the db metrics in postgres instances overview page
Kindly let us know how to troubleshoot this kind of issues and work towards resolution
This is usually a firewall issue…QAN works by making a TCP connection from Client to Server on 443 while the remaining exports make a TCP connection on TCP ports 4200x (where X can be 1- the number of exporters you’re running on the same host). So the fact that QAN is working says client can see server but no exporter data means server cannot see client. What confuses me though is that you mention CloudSQL which we don’t formally have an exporter for (for AWS readers it’s the equivalent of RDS) and so can only be monitored externally polling directly from the postgres port (typically 5432).
When I see issues like this I tend to look at https://pmm-server/prometheus/targets/ and click the “Unhealthy” to filter out the noise…you’ll usually see timeouts or deadline exceed (which is still a timeout). No matter what though, once you reestablish communication from PMM server to the target data will just start flowing again (within seconds).
Thanks for the info
Its neither port nor firewall issueThe Metrics started populating for a while and then stopped populatingKindly let us know where can we check the log and why are the exporters DOWN and how to bring them back online
What are the check boxes we need to tick when adding the remote instance, we selected use pg_Statements and SKIP TLS options
What PMM version is this in particular ? When you use agent-less deployment when exporters are actually ran at PMM Server side and if they are down it looks like they failed for some reason
We have installed PMM-2 Docker image on an on-premise machine doing a tunnel from the on prem machine to cloudsql instance
Does the Metrics Collection Parameter have impact on pushing them to Dash board
keeping them to high is the reason for causing the exporters fail to push the metrics