C4GL programs hang intermittently - platform issue?

I'm trying to troubleshoot a really odd problem at a customer site, and would appreciate any advice. First, I'll admit that its been a few years since I've been on CDI, and its great to recognize many of the same people - very fun. Anyway. System: an IBM 4-way 650 with 6 gigs of RAM running AIX 5.2 Database: 7.31.UD7 C4GL: 7.30.UC7 ESQL: CSDK 2.81 - 9.53.UC3 The data is stored on a FAStT module Very occasionally, under high load (well under machine capacity, but it only shows up at current peaks), a 4GL program will appear to "hang." It doesn't spin taking additional CPU time, it just never comes back from its "happpy place." Wherever that is. The specific program will vary, as will the specific SQL statement on which they hang. One commonality seems to be that an onstat -g ses will only show a partial SQL statement, as if it was cut off partway through. At this point I'm wondering if its an odd mismatch problem. Unfortunately the customer doesn't have a solid reproducible test case, which makes tracking this down rather difficult. I localized the last freeze down to a specific SQL, and tried a variety of permutations of it in tight loops, hitting the indexes and iterating through the data - nothing happened. Any suggestions?

rjstanford@gmail.com wrote: > I'm trying to troubleshoot a really odd problem at a customer site, and > would appreciate any advice. First, I'll admit that its been a few > years since I've been on CDI, and its great to recognize many of the > same people - very fun. Anyway. > > System: an IBM 4-way 650 with 6 gigs of RAM running AIX 5.2 > Database: 7.31.UD7 > C4GL: 7.30.UC7 > ESQL: CSDK 2.81 - 9.53.UC3 > The data is stored on a FAStT module > > Very occasionally, under high load (well under machine capacity, but it > only shows up at current peaks), a 4GL program will appear to "hang." client side > It doesn't spin taking additional CPU time, it just never comes back > from its "happpy place." Wherever that is. > > The specific program will vary, as will the specific SQL statement on > which they hang. One commonality seems to be that an onstat -g ses > will only show a partial SQL statement, as if it was cut off partway > through. > > At this point I'm wondering if its an odd mismatch problem. > Unfortunately the customer doesn't have a solid reproducible test case, > which makes tracking this down rather difficult. I localized the last > freeze down to a specific SQL, and tried a variety of permutations of > it in tight loops, hitting the indexes and iterating through the data - > nothing happened. > > Any suggestions? > There was a change in the power4 architecture which triggered some "hang" situations. Only occurs when using ipcshm. On server side problem is solved in IDS 7.31.UD1, so you should be safe here, but on the client side I think that 4GL 7.30.UC7 still have the problem. Latest 4GL version is 4.32.UC2. Since the problem is only related to ipcshm communication, you can get around it using ipcstr or soctcp.