Created attachment 1710266 [details] ovn-controller logs Description of problem: The ovn controller restarts, but previously has this error 2020-08-03T18:46:21.721Z|00103|util|EMER|lib/ovsdb-idl.c:4612: assertion row->new_datum != NULL failed in ovsdb_idl_txn_write__() Version-Release number of selected component (if applicable): How reproducible: It happens from time to time in OVN-Kubernetes CI jobs. Steps to Reproduce: maybe is a red herring, but there is a Warning before that may cause this? 2020-08-03T18:44:22.247326319Z stdout F 2020-08-03T18:44:22.246Z|00090|lflow|WARN|error parsing match "((ct.new && !ct.est) || (!ct.new && ct.est && !ct.rpl && ct_label.blocked == 1)) && (ip4.src == {$a10956707444534956691, $a13122364957363372530, $a14245307639866612073, $a15617139200530899851, $a16235039932615691331, $a17794588778302438979, $a18363165982804349389, $a4433314167141470080, $a5154718082306775057, $a5270369249448027068, $a5675285926127865604, $a6536697762898383367, $a6937002112706621489, $a9320209671274442397} && outport == @a10019124622592575031)": Syntax error at `$a6536697762898383367' expecting address set name. Actual results: Expected results: ovn-controller to handle the error and not restarting Additional info: it's more likely a problem with ovn-kube order of operations, but ovn should be more resilient and not restart.
The main problem here was that after restart, the ovnkube-node was not able to recover because the restart script deletes the ovn-remote field, thus the controller was not able to connect to the southdb https://github.com/ovn-org/ovn-kubernetes/pull/1667