[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference pamsrc::objectbroker_development

Title:	ObjectBroker Development - BEA Systems' CORBA
Notice:	See note 2 for kit locations; note 4 for training
Moderator:	RECV::GUMBELd

Created:	Thu Dec 27 1990
Last Modified:	Fri Jun 06 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	2482
Total number of notes:	13057

2454.0. "NT crush problem" by JIT081::USHIYAMA (making plans for Digital) Mon Mar 17 1997 06:11

	Hi,
	I was asked by my co-worker to solve NT crush problem with OBB.

	Environment is : 

	Server - Windows NT Alpha V3.51
		 ObjectBroker V2.5B-08

	Client - Windows NT Intel V3.51
		 ObjectBroker V2.5B-08

	Client program is using Visual Basic for UI, and OBB modules are
	in DLL. When client invoke simple method to server and the method
	return to client, NT crushes.
	On server side, the method seems to finish with no problem.

	IDL definition is something like -

module xxx {
    typedef struct aaa {
	string<21>	a;
	string<71>	b;
	unsigned long	c;
	unsigned long 	d;
    } aaa;
    typedef sequence<aaa> bbb;

    interface yyy {
	void zzz(out bbb result);
    };
};

	Curiously, when I tried to get trace log file at client side,
	crush never occurs!

	Does anyone have such experience?
	Is there a relation with crush problem and trace logging?

Regards,
Katsuhiko

T.R	Title	User	Personal Name	Date	Lines
2454.1		SEND::SLAVIN		`Tue Mar 18 1997 14:26`	4
	You may want to upgrade to a supported version of OBB - The most recent version is OBB V2.7-11 available from the CSC in Atlanta. From the SSB you can get OBB V2.7-10.
2454.2		JIT081::USHIYAMA	making plans for Digital	`Wed Mar 19 1997 00:39`	16
	>You may want to upgrade to a supported version of OBB - The most >recent version is OBB V2.7-11 available from the CSC in Atlanta. From >the SSB you can get OBB V2.7-10. I suggested my co-worker to upgrade, but he said that upgrading is difficult for customer's reason. He also said he tried upgrading to V2.6, but the application cannot be run. He said he will offer adding trace logging as workaround, and he want some reasonable explanation about this workaround. Is there possible reason that trace logging solves crush problem? Regards, Katsuhiko
2454.3	be warend ... to upgrade to ...	EMNTAL::STADELMANN	Sepp @ZUO 760-2609	`Wed Mar 19 1997 08:28`	43
	Regarding your words " .... you should upgrade to .... " Do you know what the impact is when you say that at large customers sites with tens of nearly independent kingdoms and a perhaps more conservative reporting system then we have at DEC, means? I guarantee you, you would feel as bad as they do, if you become the job to deploy OBB Middleware among an enterprise, with various heterogeneous platforms/mainframes and versions and many politically independent kingdom. They arn't as flexible as you require them to be if you claim ... to upgrade to ... If you claim for an upgrade while a customer is still in a very early evaluation cyle, then this is no problem. But once they are in an operative cycle, every claim for an upgrade makes them just head sick. Well, I was at Telecom yesterday, where some managers are very concerned about all the upgrades required. They have OBB 2.5a on MVS, not supporting C++, they have OBB 2.5b on Solaris and they have sites evaluating 2.6 and 2.7 with C++ on various platforms some Digital too. I say it in short: They realy feel bad, it's a mess, this absent of backward compatibility, this incompatibility of object references between versions, this nasty tricks to make it work toward the MVS OBB. Do your realise what it really means when you say .... to upgrade to .... it's just a planing and agreement period of about a year involved with challenges, and one has to overcome many difficulties when you run into the other kingdoms territory. A year full of frustration; A year during which OBB engineering may come up again with annother non backward compatible version to which they should for just another reason ... .... to upgrade to .... Again; They are just sad about that. Be warned. Sepp,
2454.4		SEND::SLAVIN		`Wed Mar 19 1997 12:30`	9
	You also have to consider the impact of a very small team supporting multiple versions of a product that are old. We just cannot do it and do development. Since OBB V2.5b we have had V2.5b-CLD, V2.6, V2.6-CLD, V2.7-10, V2.7-11. We also have 2 versions under development. The customer is behind in versions and many problems have been corrected in the subsequent versions, so it would be to their advantage to upgrade to a supported version.
2454.5	Customer needs explanation	JIT081::USHIYAMA	making plans for Digital	`Fri Mar 21 1997 08:35`	11
	I will not discuss support issue here. All I want to ask is : 1. Has such crush problem been already reported under V2.5B-08? 2. Is there possible reason that trace logging solve this problem? Any suggestions will be appereciated. Regards, Katsuhiko
2454.6		REQUE::BOWER	Peter Bower, ObjectBroker	`Tue Apr 01 1997 10:42`	2
	See note 1227 in the objectbroker notes conference.
2454.7	upgrading is the only answer?	JIT081::USHIYAMA	making plans for Digital	`Wed Apr 02 1997 00:53`	13
	> See note 1227 in the objectbroker notes conference. Tsutagi-san, who asked about this problem in the note 1227 of objectbroker conference, is the 'co-worker' I mentioned in base note. He asked for help to me because he couldn't solve this problem. You mean you're not sure why this problem occurs and upgrade to v2.6 is the only answer? If so, I won't ask further question. Regards, Katsuhiko
2454.8		REQUE::BOWER	Peter Bower, ObjectBroker	`Wed Apr 02 1997 12:28`	8
	> You mean you're not sure why this problem occurs and upgrade to > v2.6 is the only answer? If so, I won't ask further question. We were never able to reproduce the problem. The customer who had the problem upgraded to V2.6 and the problem went away. Therefore, the ipmt case was closed and we did not investigate further why NT crashed.
2454.9	thanks	JIT081::USHIYAMA	making plans for Digital	`Thu Apr 03 1997 00:56`	0