Operating System: All
Platform: All

+Tom, if you want to chime in here:

I'm leaning towards implementing the solution to suppress all open TransactionChains that are only x1 and only start showing x2+ from trace:transactions... but ONLY for TransactionChain, NOT for the Read[Write]Transaction! Maybe with a -a kind of CLI flag, to be able to see "all", if you really want to. Thoughts, approvals, objections?

OPNFLWPLUG-982 seems to say that there are cases where we should ignore and therefore be able to whitelist some known false positives (perhaps by stack trace content, just via String contains) ... so perhaps trace:transactions could have new arguments to ignore <N "leaks" or containing Strings.

https://git.opendaylight.org/gerrit/#/c/72362/ is adding sorting, which helps in spotting the bad from the harmless more easily.

This has come up in NETVIRT-1089 again... it's hard for a fully automated (CSIT) test to programmatically know / distinguish when there is a real problem.

I'll therefore add a new option to that CLI command to suppress leaks with less than N (option) open TXs.

https://git.opendaylight.org/gerrit/#/c/74652/ should do the trick... Once that is merged, jluhrsen can use it in NETVIRT-1089 with some "sensible value" (7 ?) for the new minOpenTransactions option to trace:transactions, and we're done here.