When using compressed indexes, the insertion of further rows
into the table may result in AF messages on secondary servers.
This occurs when a failover from the primary to a secondary
server has occurred. The AF may be seen on the old primary or
another secondary server such as RSS. Examples of the AF
messages that may be seen:
Assert Failed: Code assertion at rsbtcs.c:964
IBM Informix Dynamic Server Version 12.10.FC7D3
Who: Session(29, informix@server2, 0, 0x44ce3988)
Thread(54, xchg_1.8, 44cab510, 9)
File: rsdebug.c Line: 3882
Results: FALSE
Action: Please notify IBM Informix Technical Support.
stack trace for pid 25008 written to /tmp/af.41e711d
See Also: /tmp/af.41e711d
Code assertion at rsbtcs.c:964
Assert Failed: Code assertion at rsbtcs.c:966
IBM Informix Dynamic Server Version 12.10.FC7D3
Who: Session(29, informix@server2, 0, 0x44ce43d8)
Thread(54, xchg_1.8, 44cab510, 9)
File: rsdebug.c Line: 3882
Results: page->pg_frcnt - size >= 0
Action: Please notify IBM Informix Technical Support.
The stack trace associated with the first AF is:
Stack for thread: 54 xchg_1.8
base: 0x00000000471ee000
len: 69632
pc: 0x0000000001680477
tos: 0x00000000471fd000
state: running
vp: 9
(oninit) afstack
(oninit) afhandler
(oninit) affail_interface
(oninit) rsam_assert
(oninit) ridl_insert_key
(oninit) btinsert
(oninit) doadditem
(oninit) plogredo
(oninit) rlogm_redo
(oninit) next_recvr
(oninit) producer_thread
(oninit) startup
The secondary server message log will also contain rollforward
errors of the form:
Rollforward of log record failed. iserrno = 172
Log Record: log = 16401, pos = 0x58134, type =
OLDRSAM:ADDITEM(28), trans = 3544396888868388879
When this problem occurs it will be necessary to recreate the
problem index pages by dropping and recreating the index. If
only an HDR server is used then it may also be possible to use
the onmode -d index command to replicate the index from the
primary to HDR server. |
