[BUG] yb-master node gets stuck in CrashLoopBackOff when tls is reconfigured

[srteam2020]

Behavior

When reconfiguring spec.tls.enabled, we find that sometimes one of the yugabyte master nodes can be stuck in the CrashLoopBackOff status forever after it restarts:

kubectl get pods:
NAME                                 READY   STATUS             RESTARTS   AGE
yb-master-0                          1/1     Running            0          3m58s
yb-master-1                          1/1     Running            0          3m58s
yb-master-2                          0/1     CrashLoopBackOff   3          86s
yb-tserver-0                         1/1     Running            0          74s
yb-tserver-1                         1/1     Running            0          77s
yb-tserver-2                         1/1     Running            0          86s
yugabyte-operator-744c956b6d-m56j2   1/1     Running            0          4m8s

The log messages from yb-master-2:

I1106 19:55:28.597328    29 catalog_manager.cc:1414] Did not find previous SysCatalogTable data on disk. Not found (yb/util/env_posix.cc:1514): Unable to load consensus metadata for tablet 00000000000000000000000000000000: /mnt/data0/yb-data/master/consensus-meta/00000000000000000000000000000000: No such file or directory (system error 2)
I1106 19:55:28.597637    29 sys_catalog.cc:288] Creating new SysCatalogTable data
E1106 19:55:28.597716    29 master.cc:276] [email protected]:7100: Unable to init master catalog manager: Already present (yb/tablet/tablet_metadata.cc:264): Unable to initialize catalog manager: Failed to initialize sys tables async: Encountered errors during system catalog initialization:
        Error on Load: Not found (yb/util/env_posix.cc:1514): Unable to load consensus metadata for tablet 00000000000000000000000000000000: /mnt/data0/yb-data/master/consensus-meta/00000000000000000000000000000000: No such file or directory (system error 2)
        Error on CreateNew: : Raft group already exists: 00000000000000000000000000000000
F1106 19:55:28.597748     1 master_main.cc:131] Already present (yb/tablet/tablet_metadata.cc:264): Unable to initialize catalog manager: Failed to initialize sys tables async: Encountered errors during system catalog initialization:
        Error on Load: Not found (yb/util/env_posix.cc:1514): Unable to load consensus metadata for tablet 00000000000000000000000000000000: /mnt/data0/yb-data/master/consensus-meta/00000000000000000000000000000000: No such file or directory (system error 2)
        Error on CreateNew: : Raft group already exists: 00000000000000000000000000000000
Fatal failure details written to /mnt/data0/yb-data/master/logs/yb-master.FATAL.details.2021-11-06T19_55_28.pid1.txt
F20211106 19:55:28 ../../src/yb/master/master_main.cc:131] Already present (yb/tablet/tablet_metadata.cc:264): Unable to initialize catalog manager: Failed to initialize sys tables async: Encountered errors during system catalog initialization:
        Error on Load: Not found (yb/util/env_posix.cc:1514): Unable to load consensus metadata for tablet 00000000000000000000000000000000: /mnt/data0/yb-data/master/consensus-meta/00000000000000000000000000000000: No such file or directory (system error 2)
        Error on CreateNew: : Raft group already exists: 00000000000000000000000000000000
    @     0x7f22c21a5a3c  yb::LogFatalHandlerSink::send()
    @     0x7f22c137e866  google::LogMessage::SendToLog()
    @     0x7f22c137be3a  google::LogMessage::Flush()
    @     0x7f22c137f529  google::LogMessageFatal::~LogMessageFatal()
    @           0x4099ac  yb::master::MasterMain()
    @     0x7f22bcfd0825  __libc_start_main
    @           0x4089c9  _start
    @              (nil)  (unknown)

*** Check failure stack trace: ***
    @     0x7f22c21a3e21  yb::(anonymous namespace)::DumpStackTraceAndExit()
    @     0x7f22c137c3dd  google::LogMessage::Fail()
    @     0x7f22c137e906  google::LogMessage::SendToLog()
    @     0x7f22c137be3a  google::LogMessage::Flush()
    @     0x7f22c137f529  google::LogMessageFatal::~LogMessageFatal()
    @           0x4099ac  yb::master::MasterMain()
    @     0x7f22bcfd0825  __libc_start_main
    @           0x4089c9  _start
    @              (nil)  (unknown)
*** Aborted at 1636228528 (unix time) try "date -d @1636228528" if you are using GNU date ***
PC: @                0x0 (unknown)
*** SIGSEGV (@0x0) received by PID 1 (TID 0x7f22cd6c52c0) from PID 0; stack trace: ***
    @     0x7f22bd961ba0 (unknown)
    @     0x7f22bcfe45a6 __GI_abort
    @     0x7f22c21a3e74  yb::(anonymous namespace)::DumpStackTraceAndExit()
    @     0x7f22c137c3dc  google::LogMessage::Fail()
    @     0x7f22c137e905  google::LogMessage::SendToLog()
    @     0x7f22c137be39  google::LogMessage::Flush()
    @     0x7f22c137f528  google::LogMessageFatal::~LogMessageFatal()
    @           0x4099ab  yb::master::MasterMain()
    @     0x7f22bcfd0825 __libc_start_main
    @           0x4089c9 _start
    @                0x0 (unknown)

It seems that yb-master node failed to find previous SysCatalogTable data, and tried to create a new one. After that it encountered an error during the initialization and aborted.

From yugabyte-operator log, it encounters reconciliation error:

2021-11-07 00:14:13.638304 I | yugabyte-k8s-operator: running command 'yb-admin get_universe_config' in YB-Master pod: yb-master-0, command: ["bash" "-c" "/home/yugabyte/bin/yb-admin --master_addresses yb-master-0.yb-masters.default.svc.cluster.local:7100,yb-master-1.yb-masters.default.svc.cluster.local:7100,yb-master-2.yb-masters.default.svc.cluster.local:7100 get_universe_config"]
{"level":"error","ts":1636244084.8385901,"logger":"controller-runtime.controller","msg":"Reconciler error","controller":"ybcluster-controller","request":"default/example-ybcluster","error":"command terminated with exit code 137","stacktrace":"..."}