mike@glisten.UUCP (Michael Wendel) (07/20/90)
We are running a wide area network of Sun systems across North America connecting over 28 locations together using Sybase. There are 8 regional locations each running a Sybase server on a SUN 4/280 system with many Sun 386i and Sparcstation clients at each location (on an ethernet). In addition there are remote locations with a Point to Point connection to the nearest region. All the regions are tied together with dedicated Point to Point lines. Our application handles distribution of data among the regions and the individual Sybase servers. There are over 125 Sun 386i clients and over 150 Sparcstation clients on the network. The end user application is written using c++, sunview and Sybase DB library, with over 200 stored procedures. The current version of Sybase used in development and production is 4.0e (a major enhancement release [bug fix]) to the 4.0 release. We installed 4.0e in February. Since 4.0e was installed, and due to growth in the use of the application, we have been experiencing 3 major types of problems with the Sybase server. Ther first problem is "infected process". This is where the sybase server is managing many connections from clients and it somehow loses track of the resources used by the connection. Sometimes we can kill it, and sometimes we have to halt the server and restart it (breaking all user connections). Usually other flakey problems occur if the infected process is killed and the server left running (no halt and restart). The second problem appears to be Server corruption of some sort with regard to stored procedures. From time to time, we will start to get errors from the server where SQL code that normally works, won't. We have determined that reloading ALL the stored procedures corrects this problem. On a few occasions, the server complains that the stored procedure is bad. Usually, the error received indicates somthing unrleated, pointing at application error. The third problem is the receipt of extraneous, premature EOF from the server during an SQL request. The application sees it as a normal completion of the SQL request and goes merrily about its business with incomplete data. Examination of the Sybase error log, shows that the error occured. We no have to monitor the log continuously to catch this condition. Sybase is looking into these problems, and its a slow process. They are difficult to catch, and when they occur, our production environment requires that we recover immediately. So we can't keep the problem online for anyone to monitor very long. Now after this long description of the problems and environment, this is what I would like to find out: 1. Has anyone on the net seen these same problems? If so, have they been resolved and how? 2. Possibly migrating to 4.01, the latest release, will fix our problem. Has anyone done the migration to 4.01 from 4.0e? If so, what was involved? 3. Can I run 4.01 of the server and use the 4.0e application front ends? dwb, dblib, isql, and so on. We found that a fix applied in 4.0e dblib did not get into 4.01 dblib, so we need to stay with 4.0e dblib in the front end (client application). Any help or advice would be appreciated. Thanks in advance. ========================================================================== Michael L. Wendel (602) 464-5744 /\ General Logistics International, Inc. /\\/ 1201 S. Alma School Rd., Suite 7550, Mesa, AZ 85210 \//\ ..!asuvax!glisten!mike or attmail!mwendel GLI \/
francois@welch.jhu.edu (Francois Schiettecatte) (07/21/90)
This is a message for Michael L. Wendel of GLI I would be grateful if you would send any extra info you may have/ are going to get on the three problems you are experiencing. We have/are experiencing the EOF problem and I think we have found out why we are getting it. If I can have an e-mail address I will post that onto you, the person who found out why is out right now. Francois e-mail francois@welchlab.welch.jhu.edu