phandle not updated on node-down, node-up scenario
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Trafodion |
Fix Committed
|
High
|
Oliver Bucaojit |
Bug Description
Ron and Zalo hit a bug while running node down/up testing. The issue is that on the node-up, the phandle that the lead TM has for the down'd node does not get updated with the newer process ID. One of the issues is that it will cause sqstop to hang when the lead tm tries to coordinate shutdown.
gdb commands used to diagnose problem:
(gdb) p &gv_tm_
$1 = (SB_Int64_Type (*)[8]) 0x6cd260
(gdb) p gv_tm_info.
$2 = {iv_phandle = {_data = {35525881947320
ip_restartTim
(gdb) p *(SB_Phandle*
$3 = {iv_flags = 0, iv_type = 2, iv_vers = 1 '\001', iv_len = 64 '@', iv_name_len = 4 '\004', ia_name = "$TM1", '\000' <repeats 27 times>, iv_nid = 1, iv_pid = 10473,
iv_verifier = 3, ia_rsv2 = {0, 0, 0}, iv_oid = 1}
(gdb) c
In sqshell -a, down and up node 1
SB_Phandle will still have the older process ID.
On sqstop, the shutdown will get stuck.
Need to trace through the code to find why this is not being refreshed.
Ron checked-in a fix for this that went in Mar 25: /review. trafodion. org/#/c/ 1387/
https:/