Chapter 2
Problem Reports

This chapter describes known problems that have been fixed or that remain open, related to the QCIC software. Each problem report includes a way to work around the problem if a way exists. The number that appears in the problem report titles is the Problem Tracking System number assigned to the report.


ATTENTION

Unless otherwise stated, fixes for the problems listed here will be considered for inclusion in a future release of the product.



2.1 Fixed Problems

The following bugs have been fixed in this release:


2.2 Open Problems


2.2.1 SCSI Buses Renamed After Upgrade (250480)

When upgrading the QCIC software to V3.4.1, the SCSI buses on a machine get allocated new names which, in turn, causes the names of tape drives and robotic libraries to be changed. This can prevent backup products from functioning as expected.

Workaround. The problem can be fixed by renaming the devices to the old names, after the upgrade of QCIC V3.4.1. Using the output of dumpconf from before the upgrade of QCIC, determine the correct names for the tape drives and libraries (also Pbays and SCSI buses), compact the existing naming database (using devctl -Z), and rename the devices that have changed names using the following command:

devctl -n{currentname} {desiredname}
If you are reading this before doing the upgrade, save a copy of the output of dumpconf. If you are reading this after the upgrade and it is not obvious what the tape devices or libraries used to be called, check the contents of the naming database before compacting it. Along with the messages in the ktlog (after the first boot with new version of QCIC) these should allow you to determine the previous names for the devices.


2.2.2 Target-Initiated SCSI Negotiation Results in Error (228537)

With CLARiiON® RAID products after a DASS panic (DASS reboots) or after a raiddown command (downloads Flare code to the DASS), a problem has been seen where the QCIC firmware cannot handle target-initiated SCSI negotiation on a reselection. The QCIC firmware sends an abort message in response, causing the diskid command to hang.

Workaround: Disable target initiated negotiation (TIN) on the CLARiiON system.


2.2.3 SCSI Starvation (248319)

System configurations can exist in which too much disk bandwidth is allocated to individual SCSI buses. A Pbay fully populated with 12 high speed drives (e.g., 9GB IBM, 18GB IBM), deep device queues (i.e., greater than 2), and multiple high priority initiators would be an example of such a non-optimal set up. The ping-poll starvation management algorithm is tuned for a heavily loaded, but not a severely overloaded SCSI bus. If you need to run a non-optimal configuration like this example, some tuning of bus utilization parameters may be needed to avoid starvation failures. As a general rule of thumb, you need not do anything unless you observe an excessive number of I/O timeout entries in your ktlog. If you do have a large number of timeout events, you need to tune the SCSI bus to run as fast as possible without hitting an unacceptable level of starvation-induced timeouts. By tuning your parameters for less bus performance, you avoid starvation under heavy load scenarios. You should keep in mind, however, that you are simultaneously sacrificing some performance for light load scenarios.

Workaround: General guidelines for tuning for starvation avoidance: