Hello,

I have any problems with  ER rootserver; the server often
crashes..about 1-2 times per day.the system is Unixware 7 and IDS 9.20.
The error is :

16:20:23  Informix Dynamic Server 2000 Version 9.20.UC3
16:20:23   Who: Session(92, informix@sco00, 0, 390625404)
               Thread(80, CDRACK_1, 17462bd8, 3)
               File: mtex.c Line: 408
16:20:23   Results: Exception Caught. Type: MT_EX_OS, Context: mem
16:20:23   Action: Please notify Informix Technical Support.
16:20:23  stack trace for pid 17198 written to
/home2/informix/tmp/af.4380c56
16:20:23   See Also: /home2/informix/tmp/af.4380c56
......
16:20:42  mtex.c, line 408, thread 80, proc id 17198, No Exception
Handler.
16:20:42  Fatal error in ADM VP at mt.c:11029
16:20:42  Unexpected virtual processor termination, pid = 17198, exit =
0x100

16:20:43  PANIC: Attempting to bring system down
16:20:43  semctl: errno = 22

Constantly the rootserver crashes by reason of CDRACK_0 or CDRACK_1.
The problem appears after I define new replicates between 2 leaf
servers connected directly to rootserver.

The architecture is here:  sco00-rootserver
                          sco01,sco40,sco42,sco43,sco44,sco45,sco46
leaf servers connected directly to sco00.

I define new replicates primary target- data are replicated from
sco40-sco46 to sco01; this replicates replicate a lot of data...30-50
row(transaction)/ min. i think that this volume of data cause the crash
because before i define this new replicates I haven't this problem.

Here is part of onconfig and af. file genereted:
--onconfig
CDR_LOGBUFFERS  16384
CDR_EVALTHREADS 1,2             # evaluator threads
(per-cpu-vp,additional)
CDR_DSLOCKWAIT  300             # DS lockwait timeout (seconds)
#CDR_QUEUEMEM    4096            # Maximum amount of memory for any CDR
queue (Kbytes)
CDR_QUEUEMEM    16384
CDR_LOGDELTA    30              # % of log space allowed in queue
memory
CDR_NUMCONNECT  100             # Expected connections per server
CDR_NIFRETRY    300             # Connection retry (seconds)
CDR_NIFCOMPRESS 5               # Link level compression (-1 never, 0
none, 9 max)
--onstat -g ath

75      1816b1f0 174614d8 2    sleeping secs: 52       3cpu
CDRNsT117
77      1807fcf8 17461a98 2    cond wait  CDRBlbslp    3cpu
CDRBLOB_0
78      182a2178 17462058 2    cond wait  CDRBlbslp    3cpu
CDRBLOB_1
79      182af1a0 17462618 2    cond wait  CDRAckslp    1cpu
CDRACK_0
*80      182bc1a0 17462bd8 2    running                 3cpu
CDRACK_1
81      182c91f0 17463198 2    cond wait  CDRDssleep   1cpu
CDRD_0
82      182d7178 17463758 2    cond wait  CDRDssleep   1cpu
CDRD_1
83      182e4178 17463d18 2    cond wait  netnorm      1cpu
CDRNr46

onstat -g stk 80 light:

Informix Dynamic Server 2000 Version 9.20.UC3   -- On-Line -- Up 1 days
05:11:08 -- 433028 Kbytes

Stack for thread: 80 CDRACK_1
base: 0x182c0018
 len:   36864
  pc: 0x0856f0eb
 tos: 0x182c7dd8
state: running
  vp: 3

0x08863d98 (*nosymtab*)0x8863d98

What can i do to avoid this problem ? Can I tuning any parameters on
onconfig file?

Also I think to define server sco01 as nonroot server and sco40..sco46
leaf servers connected to sco01..is it a good idea ? My expectation is
this architecture avoid replication from sco40-46 to sco01 through
sco00 so the data will be replicated directly and sco00 won't be
implicated...is it correct ?

Thank you in advance...
Cristian