A PowerPatch 5 bug put 3000 shipments on hold

October 1998

Click here for Forsythe sponsor message

A PowerPatch 5 bug put 3000 shipments on hold

HP had to hold up shipments of new HP 3000s for about a week in early October because of one buggy patch that’s part of PowerPatch 5 for MPE/iX 5.5. The bug was serious enough to prompt HP to release PowerPatch 6 for MPE/iX 5.5 — and to recommend to its customers that they do not install PowerPatch 5. The problem can scramble datasets in IMAGE databases, causing problems not even third party utilities like Adager can fix.

In messages posted over the Internet on October 7 and 8, customers and developers began to discuss a production hold that one customer experienced when trying to order a new HP 3000. Within a day, Jon Cohen of the HP 3000 division posted instructions on the problem and notice of HP’s plans to fix the bug with a patch and new PowerPatch. By releasing the new PowerPatch 6 in mid-October, HP has made PowerPatch 5 one of the shortest-lived PowerPatches in MPE’s long history: about 60 days of useful lifespan. A single patch to fix the problem, MPEKXJ3, should be available via the HP Electronic Support Center Web site. HP was promising it expected to ship PowerPatch 6, which will include the MPEKXJ3 fix, by the week of October 19.

The problem patch within PowerPatch 5 is MPEKX79, but others superceded MPEKX79; if you have any of these patches, you are exposed to the MPEKX79 problem — MPEKXE4, MPEKXE9, MPEKX94, and MPEKXG6. MPEKX79 was supposed to be a performance enhancement patch intended to improve the file system performance when a file returns unused file space. This file system function is most typically used by a DBUTIL ERASE command, an IMAGE database utility that’s part of MPE/iX. HP’s Cohen posted the following explanation to the Internet:

“Intermittently, following an ERASE command in DBUTIL, data is not being completely erased from datasets. This results, for details, in IMAGE error -3. For master datasets, DBPUTs fail with DUPLICATE KEY VALUE IN MASTER (error 43). While the vast majority of the problem reports received stem from DBUTIL ERASE, there are a few reports of other usages hitting the same problem.” One customer said that a Cognos’ PowerHouse Quiz subfile which is erased and re-loaded will sometimes contain blank lines in the middle of the file.

High on list of things to avoid is using this DBUTIL ERASE command if a system is running on PowerPatch 5. One workaround is to use Adager’s ERASE DATASET function instead. Adager goes through every record and actually writes zeroes to it, explicitly, without depending on any file-system magic. Adager’s Alfredo Rego took note that not even Adager’s fix capabilities could save datasets that had been jumbled by the fundamental MPE/iX bug. “Unfortunately, if you trusted that DBUTIL/ERASE had, indeed, erased every dataset in your database, and if you then added new entries to datasets that were veritable Swiss cheeses (full of holes and random bits of matter), all bets are off,” he said. “The only way to ‘recover’ from this mess is to restart from scratch.”

Adager can examine every path in a database to discover which datasets have garbaged data, Rego added. “You can try to repair chains and/or to do other therapeutic operations,” he said, “which could include — as a radical amputation step — totally erasing the offending datasets so that you can begin with a clean slate.”

Some users reported that they have had PowerPatch 5, including the offending MPEKX79, installed and running in production for weeks without experiencing a problem. But Cohen was explicit about not installing this PowerPatch if managers have not already done so yet. Instead, managers are instructed to install the standalone fix patch — which takes out the MPEKX79 capabilities — or PowerPatch 6, once HP makes them available. Another option is to install MPE/iX 6.0, which never got the buggy MPEKX79 capabilities.

The problem came at a bad time for the 3000 division, holding up for a brief time shipment of new HP 3000s in the last month of the final quarter of the 1998 fiscal year. HP couldn’t ship any new systems with PowerPatch 5 because of the bug, and wouldn’t ship with PowerPatch 4 because that release didn’t support new disk drives bundled with some 3000s. Some customers were complaining about the production hold delaying their installations. HP rarely holds up system shipments in this way, but its view is that a 3000 without a bug-free bundled OS isn’t really a shippable product. Few MPE/iX releases have been completely free of bugs, however, so the production hold is some indication of the severity of this problem.
Copyright 1998 The 3000 NewsWire. All rights reserved




		Click here for Forsythe sponsor message
		A PowerPatch 5 bug put 3000 shipments on hold HP had to hold up shipments of new HP 3000s for about a week in early October because of one buggy patch that’s part of PowerPatch 5 for MPE/iX 5.5. The bug was serious enough to prompt HP to release PowerPatch 6 for MPE/iX 5.5 — and to recommend to its customers that they do not install PowerPatch 5. The problem can scramble datasets in IMAGE databases, causing problems not even third party utilities like Adager can fix. In messages posted over the Internet on October 7 and 8, customers and developers began to discuss a production hold that one customer experienced when trying to order a new HP 3000. Within a day, Jon Cohen of the HP 3000 division posted instructions on the problem and notice of HP’s plans to fix the bug with a patch and new PowerPatch. By releasing the new PowerPatch 6 in mid-October, HP has made PowerPatch 5 one of the shortest-lived PowerPatches in MPE’s long history: about 60 days of useful lifespan. A single patch to fix the problem, MPEKXJ3, should be available via the HP Electronic Support Center Web site. HP was promising it expected to ship PowerPatch 6, which will include the MPEKXJ3 fix, by the week of October 19. The problem patch within PowerPatch 5 is MPEKX79, but others superceded MPEKX79; if you have any of these patches, you are exposed to the MPEKX79 problem — MPEKXE4, MPEKXE9, MPEKX94, and MPEKXG6. MPEKX79 was supposed to be a performance enhancement patch intended to improve the file system performance when a file returns unused file space. This file system function is most typically used by a DBUTIL ERASE command, an IMAGE database utility that’s part of MPE/iX. HP’s Cohen posted the following explanation to the Internet: “Intermittently, following an ERASE command in DBUTIL, data is not being completely erased from datasets. This results, for details, in IMAGE error -3. For master datasets, DBPUTs fail with DUPLICATE KEY VALUE IN MASTER (error 43). While the vast majority of the problem reports received stem from DBUTIL ERASE, there are a few reports of other usages hitting the same problem.” One customer said that a Cognos’ PowerHouse Quiz subfile which is erased and re-loaded will sometimes contain blank lines in the middle of the file. High on list of things to avoid is using this DBUTIL ERASE command if a system is running on PowerPatch 5. One workaround is to use Adager’s ERASE DATASET function instead. Adager goes through every record and actually writes zeroes to it, explicitly, without depending on any file-system magic. Adager’s Alfredo Rego took note that not even Adager’s fix capabilities could save datasets that had been jumbled by the fundamental MPE/iX bug. “Unfortunately, if you trusted that DBUTIL/ERASE had, indeed, erased every dataset in your database, and if you then added new entries to datasets that were veritable Swiss cheeses (full of holes and random bits of matter), all bets are off,” he said. “The only way to ‘recover’ from this mess is to restart from scratch.” Adager can examine every path in a database to discover which datasets have garbaged data, Rego added. “You can try to repair chains and/or to do other therapeutic operations,” he said, “which could include — as a radical amputation step — totally erasing the offending datasets so that you can begin with a clean slate.” Some users reported that they have had PowerPatch 5, including the offending MPEKX79, installed and running in production for weeks without experiencing a problem. But Cohen was explicit about not installing this PowerPatch if managers have not already done so yet. Instead, managers are instructed to install the standalone fix patch — which takes out the MPEKX79 capabilities — or PowerPatch 6, once HP makes them available. Another option is to install MPE/iX 6.0, which never got the buggy MPEKX79 capabilities. The problem came at a bad time for the 3000 division, holding up for a brief time shipment of new HP 3000s in the last month of the final quarter of the 1998 fiscal year. HP couldn’t ship any new systems with PowerPatch 5 because of the bug, and wouldn’t ship with PowerPatch 4 because that release didn’t support new disk drives bundled with some 3000s. Some customers were complaining about the production hold delaying their installations. HP rarely holds up system shipments in this way, but its view is that a 3000 without a bug-free bundled OS isn’t really a shippable product. Few MPE/iX releases have been completely free of bugs, however, so the production hold is some indication of the severity of this problem. Copyright 1998 The 3000 NewsWire. All rights reserved