- There was a M4.0 event in Orange County at about 16:20 on Monday. As
always, the traffic on the web server started increasing within two minutes,
and peaked at 82 hits/second after 10 minutes. This was notable because
this was higher than the 67 hits/second we had after Hector Mine. The
squid server handled this traffic easily.
- Swapped the ethernet card in the USGS squid server. Reading the comments
by the author of the driver for the RealTek card showed that the card we
were using was inefficient, and generally not a good performer under high
loads. This explains why the server reached 100% CPU usage in testing at a
data rate of about 65Mb/sec. Got a 3Com 3C905TX to replace it. This should
allow the server to run all the way up to the bandwidth limit of about
90Mb/sec.
- The disk in channel 0, target 5 in the Spring AA RAID failed. Replaced
it with a cold spare. Got RMA #20824 from nStor and got it packaged up to
send back.
- Repartitioned the system disk space on Hotspot to fix the disk layout.
This was required because the ISTI people wanted to install gcc on the
system, and there was not enough space left on /usr to allow for this
installation. This disk looked like it was set up by the Solaris
auto-layout sytstem, which always makes the system partitions too small.
- Got another disk back from nStor. Put this one in the AA RAID on Spring
to act as the hot spare. Now both AA RAIDs have a hot spare.
- Hooked up a serial cable from Hotspot to the former Contel paging modem
so that Patrick could use the modem for testing paging.
- The Simpson map on Ehzsouth was out of date on Friday morning. This
turned out to have been caused by QDDS dying for some unknown reason.
After restarting it, the map updated correctly.
- Put patch 105181-19 on Granite to fix the kernel to allow it to use the
faster CPU in the Ultra-10. Did the hardware swap again, and the new
system works properly.
- The USGS Squid server rebooted mysteriously at 19:20 on Monday.
- Examining the logs for the USGS squid server showed that people have felt
and come to our web site after events as small as M1.5. The graphs in the
daily usage reports display a characteristic curve after a felt event.
- Fixed the Big Brother clients on Usgs-squid and Eqinfo to have them send
status reports to both Big Brother display machines.
- Repartitioned the disk on Onyx to make larger /usr and /var partitions.
The history of system patches had eaten up all the /var space.
- Talked to Larry Watanable from ITS about what we need to do to move the
ITS networking equipment and our computers out of the Yellow House. Also
arranged with the Electrical Shop people to have them put in power and
network wiring in the alcove by the kitchen in the Tan House.
- Hooked up the Anaheim packet radio to Hotspot for Patrick to use in
testing the new paging programs.
- Built the new Trinet-squid server.
- Installed the Trinet-squid server on a 100Mb port in the rack next to
Flint.
- Assisted with moving the SCIGN computer racks to the Tan House.
- Arranged with ITS to move the Catalyst 5000 into the Tan House basement
in preparation for the remodel.
- Rebooted wwolf to clear a networking problem. Network services on wwolf
were all unavailable for unknown reasons.
- Configured the SLIP server for GPO. At first, the SLIP lines would not
come up. This turned out to have been caused by the server having AUTOBAUD
enabled on those lines. Disabling this feature fixed the problem.
- Carizo experienced a problem with high CPU load which we traced to it
having a high level of interrupt activity. This turned out to be a bug in
the new version of UCX and its implementation of telnet. Incoming telnet
connections from unix systems were unable to negotiate the terminal type,
and so the telnet receiver on Carizo was chasing its tail trying to
renegotiate with the unix client. Setting the TERM environment variable to
"vt100" is a workaround for this problem.
- Moved John Galetzka's images and imgbank directories off the /webd1 disk
on ehzsouth. There was not enough space on the disk.
- Got two replacement disks for the RAIDs back from nStor. Put one as
standby in the Spring SX. The other one is a cold spare for the AA RAIDs.
- Carizo died mysteriously at 4 minutes after midnight on the morning of the
30th. It acted like it did not have power, but after taking it apart to
check the power supply, it acted normal and booted up.
- Attended the CUBE users meeting.
- The Eqinfo mailing list server went live on the 30th. The subscription
page was put up on the main Pasadena web page, and people started signing
up.
***
- Set up connections to Hotspot for dialout and packet radio paging testing.
- Built and configured the new squid server for www.trinet.org/. This will
increase the load capacity of the Trinet web site.
- Arranged with ITS and Physical Plant to get the necessary network and
electrical connections set up for the computers that will have to be
relocated during the Yellow House remodeling.