QV Peers – I’m looking for some troubleshooting advice, recommendations, clarification, and advice on this issue. My environment is QV 12.10 SR6, QVS cluster with 2-nodes [Loaded Document load balancing] (QVDS, QMC, QVWS all on separate nodes as well), QVS Event Log Verbosity = High.
Issue: Over the last 3 months I’ve noticed that about once a week in the QMC, QVS Statistics shows that one of our QVS nodes has no documents loaded, while the other has ~15 documents loaded. QMC Status->Services shows both QVS nodes as “green”, and no users are reporting any problems. In AccessPoint I select an application which is not loaded and load it – it loads on the “good” QVS node, and not on the QVS node with zero apps loaded. The server is not showing any issues – memory consumption is around half, and CPU is idle. I restart the QVS on the “bad node” and notice that applications are now loading on both nodes. No Errors are recorded in the logs. I’m puzzled – this appears to have been a hung QVS, but Qlik did not indicate it as such.
Can any of you clarify:
- When QVS Statistics loads, can it time out? That is can the QMC show that no apps are loaded on the QVS Node erroneously because the query between the QMC and the QVS node timed out? When QVS Statistics shows no apps loaded – can I actually trust that?
- When QMC Services shows that QVS is “green” – what does that actually mean? Is that the same as the Windows OS showing the QVS service as “running” or does the QMC attempt to “connect” to the QVS for a deeper level of interrogation? If the status is “green” – can I trust that, or might the QVS still be hung up and show as green?
- Anyone else experience issues where a QVS node is hung up, but the QMC doesn’t indicate issues?
- Under the “Loaded Document” load balancing algorithm, when QVWS is deciding which QVS node to utilize – how does it do this (beyond the help documentation explanation)? Does it run a query against the QVS to see what’s loaded (and is there a way I can run this same query)? Might this query timeout causing the load balance algorithm to suddenly favor a specific node (and can this timeout be adjusted in a settings file)?
- If a specific QV application is hanging my QVS node – what do I look for in the logs? Excessive memory utilization? Do I look for apps unloading as they “fail over” to the other node? How might I pinpoint the problematic application (we have ~150 apps)?
- If the QVS is hung why would QMC show it green?
Any other advice, troubleshooting tips, or things to consider? (FYI – I have previously implemented changes suggested by QlikSupport but it did not resolve my issue).
Thanks in advance for any input!