Re: Undefined replicate crashes IDS

Yes - this is probably bug 155787. I'm really sticking my neck out on offering this workarround, so I need you to be very careful. Otherwise, you will be forced to delete ER on all of your nodes and re-establish it. You might be able to get arround the problem by setting the ATS/RIS spooling directories to spaces. This is done by... dbaccess syscdr <<!EOF update servdef set atsdir = "", risdir = "";!EOF However, I'm not sure if the failure is occuring before or after the check is made to see if the directory has been set. If a failure still occurs, then you will need to turn off ATS/RIS spooling on the replicate. We need to turn off the two bits that indicate ATS/RIS spooling from the replicates. The two bits that need to be turned off are RIS (0x400) and ATS (0x800) from the replicate flags. I'm enclosing a procedure that should work to turn off the bits. These two bits have the value of 3072. The following procedure should be able to turn off these bits from the flags. However, it will not work if the high order bit is set (0x80000000) which are actually negative values. So by creating the following procedure and executing the following while CDRBLOCKOUT is set, you should be able to turn off the two bits. This must be run as user informix. dbaccess syscdr <<!EOF update repdef set flags = bitoff(flags(3072));!EOF Once that is done, then you could try to bounce the server with CDRBLOCKOUT is turned off. N.B. - like I said. I am really sticking my neck out on this, so you will need to be prepared to rebuild ER if this doesn't work. ----------------------------------------------------------------------- create procedure bitoff(flag int, bitset int) returning int;define result int; define tflag int; define mask int; define tmask int; define tcnt int; define t1 int; define returnval int; let result = 0; let tmask = bitset; let tflag = flag; let mask = 1; if (flag < 0) then return flag; end if if (bitset < 0) then return flag; end if for tcnt = 2 to 31 let t1 = mod(tmask, 2); let tmask = tmask / 2; if (t1 = 1) then let t1 = mod(tflag,2); if (t1 = 1) then let result = result + mask; end if end if let tflag = tflag / 2; let mask = mask * 2; end for if (tmask != 0) then if (tflag != 0) then let result = result + mask; end if end if if (bitset > 107374183) then if (flag > 107374183) then let result = result + 1073741824; end if end if return flag - result; end procedure; ----------------------------------------------------- "Denis Melnikov" <dm'Inik@r'gent.r'> wrote in message news:btdnae$es8$1@n6.co.ru... > Is it what you need? > > ===========------------- - - - - - - > /usr/informix/server/bin/onstat -g stk 20852 light: > > Informix Dynamic Server 2000 Version 9.21.UC2 -- On-Line -- Up 16 days > 00:24 > 52 -- 1654784 Kbytes > > Stack for thread: 20852 CDRD_9 > base: 0x96b37018 > len: 66048 > pc: 0x085138eb > tos: 0x96b45fd4> state: running > vp: 4 > > 0x0852b06d (oninit)afhandler(0x96b460fc, 0xb, 0x969156a0) > 0x0852aa6a (oninit)afcrash_interface(0x3, 0x8749ef0, 0x96b46574) > 0x0852d57f (oninit)mt_ex_throw_sig(0x8749ef0, 0x96b46574, 0x96d1b648) > 0x0850ca0e (oninit)afsig_handler(0x1, 0x8746534, 0x0) > 0xdfbbd32f (*nosymtab*)0xdfbbd32f(0xb, 0x96b46960, 0x96b46760) > 0xdfbca441 (*nosymtab*)0xdfbca441(0xb, 0x96b46960, 0x96b46760) > 0xffffffff (*nosymtab*)0xffffffff(0xb, 0x96b46960, 0x96b46760) > 0x085df288 (oninit)dsiSpool(0x6f, 0x96346708, 0x95c66020) > 0x085da429 (oninit)dsiAbortTx(0x95c66020, 0x96b47080, 0x96b470f0) > 0x085d999a (oninit)dsiResolverThread(0x95f2e0f8, 0x96346ef8, 0x96b47080) > 0x085153a3 (oninit)startup (0x96346ef8, 0x0, 0x0) > 0x00000000 (*nosymtab*)0x0 > > > "Madison Pruet" <mpruet@comcast.net> '''''''/'''''''' ' '''''''' ''''''''': > news:xzgKb.741616$HS4.5686487@attbi_s01... > > 9.21 uses the environment variable CDRBLOCKOUT to start the engine without > > ER. I'd like to see the stack trace. > > > > There was a problem with ER trying to print an ATS file when it could not > > get to the description of the replicated row. I'm guessing that is > causing > > the problem, but would need to see the stack to determine that. > >

Thank you, Madison, for your assistance! I succeeded to bounce the engine with CDRBLOCKOUT set/unset, and IDS started up successfully. So I haven't used your workaround, but thank you anyway. It was in such way: since the dest server have brought up with ER, the src server run CDRNrA9 thread (CDR receiver, I think). It occupied the whole CPU for a long time (~15 min.), so I thought it will never end, and bounce the src server. (I determine running threads with 'onstat -g wst'). When the src server started up, it run CDRNr9 thread (receiver - yeh? - again). It worked ~15 min. as well. And then CDRNr9 changed with CDRNsT9 (sender?) as well as ddr_log_io threads. Since that moment all spooled transactions were replicated in 25 min. Thank you again!