GI 19.31 and 19.32: trace file flooding in mgmtca.trc can happen

Thanks to Peter Lehmann who always alerts me about findings and encounters he had in his test and production RAC cluster environments. Peter dropped me an email the other day, and I share his suspicion that this may effect many in the coming weeks operating RAC environments. If you don’t have a RAC, you may not see this happening with GI 19.31 and 19.32: trace file flooding in mgmtca.trc can happen.

Photo by Giovanna Gomes on Unsplash

What happens?

It took me a bit of searching, and Peter collected a lot of information already.

Two months after patching GI to 19.31 he discovered flooding the trace mgmtca.trc on their RAC environments. This is the message he found in the mgmtca.trc:

[mydb] [ 2026-08-28 14:47:58.971 CDT ] [Version$VersionEnum.getEnumMember:264]  Version Match Successful: returning version object 19.0.0.0
[mydb] [ 2026-08-28 14:47:58.971 CDT ] [MgmtcaCommandLineParser.parseSetPasswd:478]  Parsing setPasswd with args:[-user, CHMOS,CALOG,QOS]
[mydb] [ 2026-08-28 14:47:58.975 CDT ] [Version.isPre:757]  version to be checked 19.0.0.0.0 major version to check against 18
[mydb] [ 2026-08-28 14:47:58.975 CDT ] [Version.isPre:768]  isPre.java: Returning FALSE
[mydb] [ 2026-08-28 14:47:58.995 CDT ] [Mgmtca.mydb:436]  MGTCA-1117 : The user name specififed in the 'mgmtca' command is invalid.
[mydb] [ 2026-08-28 14:47:58.996 CDT ] [Mgmtca.mydb:442]  MGTCA-1117 : The user name specififed in the 'mgmtca' command is invalid.
oracle.cluster.mgmtcommon.MgmtDBException: MGTCA-1117 : The user name specififed in the 'mgmtca' command is invalid.
	   at oracle.cluster.mgmtca.MgmtcaCommandLineParser.parseSetPasswd(MgmtcaCommandLineParser.java:492)
	   at oracle.cluster.mgmtca.MgmtcaCommandLineParser.parse(MgmtcaCommandLineParser.java:1377)
	   at oracle.cluster.mgmtca.Mgmtca.execute(Mgmtca.java:690)
	   at oracle.cluster.mgmtca.Mgmtca.mydb(Mgmtca.java:429)
[mydb] [ 2026-08-28 14:48:18.127 CDT ] [SRVMContext.init:205]  Performing SRVM Context init. Init Counter=3
[mydb] [ 2026-08-28 14:48:18.132 CDT ] [has.HASContextNative.Native]  prsr_trace: Native: allocHASContext

< --- cut here --->

Furthermore, the ologgerd.trc shows:

2026-08-28 14:48:59.019 : CRFLOGD:3488605952: crflogdb_put: Insertion failure. dbxret[-11].
2026-08-28 14:48:59.019 : CRFLOGD:3488605952: LDBWRITE: Hit a retriable failure in insertion for the 25165th time. ret[-11]. Retrying the insertion now...
2026-08-28 14:48:59.019 : default:3488605952: crflogdbora_put: Creating insert connection 1
2026-08-28 14:48:59.025 :  CRFOCI:3488605952: crfoci_conn_create: Connection string discoverd
2026-08-28 14:48:59.045 :  CRFOCI:3488605952: crfoci_conn_create: Trying to connect to (DESCRIPTION = (CONNECT_TIMEOUT=60) (ADDRESS = (PROTOCOL = tcp)(HOST = 169.254.9.138)(PORT = 1525))(CONNECT_DATA = (SERVICE_NAME = gimr_xxxxx_20))(SECURITY=(ENCRYPTION_CLIENT=requested))).
2026-08-28 14:48:59.079 :  CRFOCI:3488605952: crfoci_checkerr: Error - OCI_ERROR ORA-28001: the password has expired

In Peter’s case, the mgmtca.trc had already grown to 8GB of size.

Why is this happening?

In short, the reason is that as part of the WLM deprecation, the QOS user is dropped. Consequently, ologgerd attempts to reset the QOS user’s password, which results in this error being reported.

I found a number of bugs with it, especially when searching for the error MGTCA-1117:

• Base bug:
  Bug 35082763 – CRF FAILS TO RECOVER PASSWORD FOR QOS USER IN MGMTDB
• Duplicat pointing to the base bug and describing the issue for 19c:
  Bug 39480336 – 19.31: MGTCA-1117 : THE USER NAME SPECIFIFED IN THE ‘MGMTCA’ COMMAND IS INVALID

Workaround or Solution?

As a workaround, you can truncate or remove the traces on a regular basis. But a more clear solution would be applying the available one-off fix for:

• Bug 35082763 – CRF FAILS TO RECOVER PASSWORD FOR QOS USER IN MGMTDB
  • Currently available on top of 19.31 on Linux OCW
  • Currently available on top of 19.32 for Linux OCW

In addition, a MOS note (KB920277) MGTCA-1117 : The user name specififed in the ‘mgmtca’ command is invalid has been published.

This MOS note mentions another workaround:

<GI_HOME>/bin/mgmtca setpasswd -allusers
<GRID_HOME>/bin/crsctl stop resource ora.crf -init
<GRID_HOME>/bin/crsctl start resource ora.crf -init

which I think will be a temporary solution only.

Update: The inclusion of this fix for 19.33 is confirmed. With combined forces it got pushed forward, and there should be no need for single-fixes “on top” from 19.33 onward anymore.

And thank you, Peter, for telling me about it!!

Further Links and Information

• (KB920277) MGTCA-1117 : The user name specififed in the ‘mgmtca’ command is invalid
• Both bugs are non-public – thus not linked here

–Mike