Instance parallel recovery stuck on RAC database (Doc ID 2189014.1)

Last updated on JANUARY 17, 2023

Applies to:

Symptoms

 RAC parallel instance recovery stuck :

You will see parallel process consuming 100% CPU on the host.

Investigation:
.
- When Inst1 was doing ALTER DATABASE OPEN, Inst4 was doing recovery for
Inst1
- On Inst4, parallel recovery was selected (32 slaves) and SMON was waiting
for slaves to complete the operation
o SMON was seen hung (Short stack dump)

ksedsts()<-ksdxfstk()<-ksdxcb()<-sspuser()<-__sighandler()<-
semtimedop()<-skgpwwait()<-ksliwat()<-kslwaitctx()<-kxfpqidqr()<-kxfpqdqr()<-kcrprd()<-kcrpst()<-kcvirv()<-ktmmon()
<-ktmSmonMain()<-ksbrdp()<-opirip()<-opidrv()<-sou2o()<-
opimai_real()<-ssthrdmain()<-main()<-__libc_start_main()
.
o The parallel slave processes were seen looping in kcrahtini()
Short stack dump:

ksedsts()<-ksdxfstk()<-ksdxcb()<-sspuser()<-__sighandler()<-kc
rahtini()<-kcrpst()<-kcrpdv()<-kxfprdp()<-opirip()<-opidrv()<-sou2o()<-opimai_real()<-ssthrdmain()<-main()<-__libc_start_main()

Cause

My Oracle Support provides customers with access to over a million knowledge articles and a vibrant support community of peers and Oracle experts.