Uploaded image for project: 'Observium'
  1. Observium
  2. OBS-5355

TP-Link JetStream: agent silent for ~30 s after max-rep 50 bulkwalk on missing table, baseline retry fails, device marked down

    XMLWordPrintable

Details

    • Vendor Bug
    • Resolution: Fixed
    • Major
    • None
    • Professional Edition
    • Discovery
    • None

    Description

      Observium: Professional 26.11.15526. Started with the update r15351 -> r15461 on 2026-11-04 (SNMP collection rework), still present on 15526.
      Devices: TP-Link TL-SG2210P (JetStream 5.0.0 Build 20211201) and SG2210P (5.20.0 / 5.20.1), os tplink-jetstream, SNMP v2c, 37 devices.

      Problem:

      • Discovery walks tables these switches do not have with snmpbulkwalk -Cr50 (adaptive max-rep): pingCtlEntry (sla), cbQosIfType (cisco-cbqos), diskIOEntry (ucd-diskio).
      • The agent does not answer and then stays silent for about 30 s, also for plain GETs.
      • The baseline retry with -Cr10 is sent right after the timeout, falls into this silent period and times out too. So the per-MIB max-rep from r15520 never applies, and the next discovery starts with -Cr50 again.
      • During the silent period the status check (snmpget -t 3 -r 1) fails and the device is marked "Down (snmp)". For us: about 25 false device-down alerts per day.

      Reproduction (one switch, same options as Observium: -t 2 -r 2, 20-30 s pause between steps)
      1. snmpbulkwalk -Cr10 ... diskIOEntry -> 0.1 s, "No Such Object"
      2. snmpbulkwalk -Cr50 ... diskIOEntry -> 6.0 s, Timeout
      directly after: snmpbulkwalk -Cr10 ... diskIOEntry -> 6.0 s, Timeout
      3. snmpbulkwalk -Cr50 ... diskIOEntry in the background, snmpget -t 1 -r 0 sysUpTime.0 every 1-2 s:
      no answer from 0 s to 28 s, answers again from 30 s
      pingCtlEntry and cbQosIfType behave the same with -Cr50 (6 s timeout); with -Cr10 alone they answer at once.

      Workaround in place: discovery_blacklist for sla, cisco-cbqos and ucd-diskio on tplink-jetstream. No false alerts in the first two full discovery runs since (before: alerts in every run).

      Suggestion: a lower default max-rep in the tplink-jetstream definition (as r15481 did for Windows), and/or remember a failed max-rep 50 walk per device/MIB even when the immediate baseline retry also fails (or delay that retry).

      Attachments

        1. benchmark-r15536.txt
          9 kB
        2. benchmark-r15537.txt
          9 kB
        3. discovery-r15536.txt
          1 kB
        4. discovery-r15537.txt
          2 kB
        5. OBS-5355-benchmark.txt
          8 kB
        6. OBS-5355-discovery-debug.txt
          161 kB
        7. OBS-5355-poller-debug.txt
          132 kB

        Activity

          People

            landy Mike Stupalov
            fake Thomas "fake" Högemann
            Votes:
            0 Vote for this issue
            Watchers:
            3 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: