IBM RS/6000 SP, Руководство по определению проблем

Страница: 137 / 316

This soft copy for use by IBM employees only.
6. -8 is the most c o m m o n indicator of a miswire. You should look for a
cable_miswire file on the primary node.
7. -9 could mean that the switch element (chip or adapter) was failing.
8. -10 through -15 are indicators that something failed during testing or a switch
chip
′
s RAM. The switch board on which it resides is the FRU.
9. -16 usually indicates some sort of clock setup problem. This is most often
reported on the nodes. For some reason the switch chips are usually
reported with a -2, -3, or -4 when they have clock problems. If you see it on
a node, what usually has happened is that the node was powered-on before
the switch had its clock set up to the correct clock. In the case of a single
switch system, the node was powered-on before the switch. There was also
an old bug where the power-on diagnostics selected the wrong clock on
which to run: a gore clock instead of a data cable clock when attached to an
HiPS 3.0, or a data cable clock instead of a gore clock when attached to an
HiPS 2.0.
10. -17 is more of a warning than an error. Make sure that it makes sense that
there is a wrap plug on that port. It makes sense when the node is not
populated or when the cable has been removed for some reason.
11. -18 is similar to -17, but occurs much more rarely.
12. -19 is very similar to -7 and was introduced later in the program when we
began distributing routes across the switch. The fault service daemon on the
receiving node must have time to service the route table information.
13. -20 is similar to -8, but is rarely seen.
14. -21 and -22 are very strange and have never been seen. Call for help!
15. -23 will be seen more often when more nodes are fenced in a system.
16. -24 should never occur. If you see it, call for help.
This file is very useful, because this is a dump of the topology file with some
comments about errors found when the switch was initialized. Looking at this
file, you can find at a glance some common wiring and node down problems.
4.7.2 Patterns in out.top File
1. Uninitialized = -1 - > look for pattern
2. Link errors = -2, -3, -4 - > look for pattern
3. Miswires = -5, -8, -20 - > check cable_miswire
4. Node time-outs = -7, -19 - > look for pattern
5. Unusual which imply FRU immediately = -6, -9 through -15 - > replace FRU
6. Clock = -16 - > look for pattern
7. Warnings = -17, -18, -23 - > verify the statement it makes is expected
8. Call for help = -21, -22, -24 - > call
Categories that require pattern recognition: 1, 2, 4, and 6.
Patterns:
1. An internal port (ports 5, 6, 7, 8) is reporting problem
Chapter 4. The Switch
117