IBM’s problems with its 3380K and 3380J disk drives are more widespread than previously reported (CI No 1,201). IBM has said that spindle bearings in its flagship disk files are failing due to an engineering problem, not as a result of some flawed components provided by suppliers. The design error affects 3380 disks worldwide; it […]
IBM’s problems with its 3380K and 3380J disk drives are more widespread than previously reported (CI No 1,201). IBM has said that spindle bearings in its flagship disk files are failing due to an engineering problem, not as a result of some flawed components provided by suppliers. The design error affects 3380 disks worldwide; it is not confined to ones made in the United States. Disk/Trend Inc, the Mountain View, California company regarded as the most authoritative independent source of information on the disk industry, reckons that in excess of 22,000 3380Ks and more than 10,000 3380Js were shipped by IBM worldwide through the third quarter of 1988, when IBM changed the way it made its head-disk assemblies. The total capacity of these machines is more than 170 trillion characters. IBM maintains that only a small number of its disks have been hit by bearing failures, and that the flaw was corrected in its manufacturing process at the end of last year’s third quarter. Most customers have not and will not encounter this situation, asserted an IBM spokeswoman. It has affected only a small number of the 3380Js and 3380Ks shipped prior to the production fix. But reports from some customer sites indicate that IBM has had to replace head-disk assemblies on portions of their disk farms ranging from 15% to nearly all the disks shipped in the middle of last year. IBM says that users get a warning before their 3380s go down. When a bearing failure occurs, it causes the spindle holding the recording media to wobble. Temporary data loss At first, the result is what IBM calls a temporary data loss. Subsequently, the problem may worsen, resulting in a permanent error. This is not, IBM insists, the same as a head crash because the data is still physically on the disk. Even though there have been some cases of data loss… in every case customers have been able to recover the data through normal data recovery procedures, IBM says. According to non-IBM sources, this recovery procedure generally involves resetting the map in a disk’s directory that shows where good and bad portions of the drive lie; this map is used to lock out error-prone portions of the recording surface. Then the drive is forced to re-read the bad spots until the data can be drawn off and copied to a fully functional disk. In an effort to help users catch failing disks before a catastrophe can occur, IBM has developed code for its 3990 and 3880 disk controllers that provides for the detection and recognition of a potential disk failure due to the bearing problem. This software, according to IBM, has been distributed to all customers with 3380J and 3380K disks. With this early warning or predictive code, explains IBM, a customer no longer gets the temporary [general] error for this situation. Rather the user and IBM’s service engineers are specifically told that a disk’s bearings are causing data errors and warned that the machinery could further deteriorate. IBM has thus far decided it will not pre-emptively replace all the head-disk assemblies it made prior to October 1988, although some sites reportedly have been able to get the company to do just this. Says IBM, all customer actions are taken on a case-by-case basis. For a temporary data check [another term for a temporary error], replacement would be foolish. Temporary data checks occur frequently for a variety of reasons. Meanwhile, users are still experiencing sporadic disk failures. The owners of the drives, which can cost in excess of $100,000 a unit, have become concerned about their investments in IBM equipment and the valuable data entrusted to it. Independent lessors, in particular, are checking with their customers to make sure any unstable disks are promptly identified and repaired. Like end users that own the drives, lessors hope to minimise the chances that they will be stuck with disks that cannot be trusted. Hesh Wiener (C) 1989 TNA Co Inc.
