Hi, Mongo Enterprise should work fine with PMM. Can you explain more about your setup? what versions are you using? Where exactly do you see this Status: unknown?
This typically means metrics aren’t reaching PMM server. To narrow this down:
In Explore, query up{node_id=“your_node_name”}. Do you see anything?
If nothing comes back, run pmm-admin status and pmm-admin list on the client. Check that node_exporter is listed and running, and that vmagent is running.
Look in the pmm-agent logs for vmagent remote-write errors, for example connection refused, 401, TLS errors, etc.
I’ve seen for one of the cases, I see that node_exported is listed and running for it.
In the logs I put AI analysing and got this:
34 exporters fail ~100% of _hr scrapes (894–902 logged failures each, error EOF or Client.Timeout exceeded while awaiting headers) — and these are exactly the replset-member exporters on the dedicated hosts. This is the same ethtool/diagnosticdata bug already diagnosed: pmm-agent.log has 16,356 panic lines / 5,453 giant_hdr lines, still panicking as of 14:30 today. _lr keeps delivering mongodb_up for all of them, which is why the services still show connected but hr-only metrics are missing / UI shows Unknown.
Looks like something is wrong with the agent connecting to mongo. Can you share the commands you used for pmm-admin add ? you can remove sensitive info like password.
Enterprise isn’t the blocker; the evidence points to the high-resolution MongoDB exporter failing while the low-resolution path still reports mongodb_up. Check that the pmm-admin add mongodb command targets the replica-set member’s advertised host/port (not localhost or a load balancer), that the monitoring user has clusterMonitor on admin, and that TLS/auth options are identical for every member. EOF/timeouts often mean the exporter can’t complete the MongoDB handshake. Also capture pmm-agent version and panic logs; the giant_hdr panic is an agent bug, not an Enterprise restriction.
Position: Your operational checklist items are all ruled out — the evidence points to the exporter bug you (Percona) yourself flagged.
Address is correct — all 34 services registered against their own member host:port; no localhost/LB/VIP.
Privileges fine — root on admin (superset of clusterMonitor).
Handshake theory refuted — the same 34 exporters that fail high-res all report mongodb_up=1 and serve the low-res fcv+pbm collectors fine (200 OK, 0.1s), which requires a fully authenticated connection. The failure is specific to the diagnosticdata collector: exporter panics server-side on mongodb_sys_ethtool_ens192_giant_hdr and drops the connection (instant EOF, reproducible 34/34, timeout-independent — tested 120s vs 4.5s, same result).
Versions + panic logs delivered — pmm-agent 3.10.0 (378dbc37), mongodb_exporter v0.45.0 (e2dd9bba), full giant_hdr stack trace at exporter.go:213: mongodb_sys_ethtool_*_giant_hdr duplicate registration in the diagnosticdata collector
after talking to eng team what triggers the bug is the operating system and network card on the host, not the MongoDB Enterprise edition.
Workaround
Re-add the affected services with pmm-admin add mongodb … --disable-collectors=diagnosticdata. This stops the panics and the log noise, but those services still won’t have diagnostic metrics until the fix for https://perconadev.atlassian.net/browse/PMM-15249 ships