- Upgraded VMS on Carizo and Ojai from 6.2 to 7.2.
- Attended a meeting to discuss options for increasing our web service
capacity. This was prompted by the failure of our present server to deal
with the traffic spike experienced after Hector Mine.
- Conference call with Menlo on web server issues. We agreed to move our
server back into the round-robin domain, but with the understanding that
the server would always direct traffic to our server, and that the DNS
would serve names with a 6 hour time-to-live, which should cut down on the
name lookup problems we had in the past. We also agreed to continue to
host a mirror of the Menlo web page on our server here.
- Began testing squid in the reverse-proxy caching mode. Using Iron as a
front-end for Flint resulted in a factor of 5 performance increase.
- Installed FreeBSD on one of the spare new PCs for use as a dedicated
squid server.
- Boron was broken into on Tuesday night. Big Brother reported an anomaly
almost immediately, since the logs were deleted. Spent the rest of the
week fixing the system and looking for the trap doors the hackers installed.
They came back every day until the trojan-horse /bin/login program they
installed was removed.
- Disabled most of the serviced in inetd.conf for the Trinet client
workstations. Also enabled TCP wrappers on Iron to log all incoming login
requests.
- ITS moved both Ehzsouth and the new squid server to 100Mb ports.
- Tested the squid server for 6 hours on Friday with a sustained load of
300-350,000 hits per hour, which is about one order of magnitude higher
traffic than we had after Hector Mine. The server performed well under
this load.
- Installed Big Brother version 1.3 on Ehzsouth.
- Turned the Calendar Manager back on for Nitya and Garnet.
- Configured an 8-port terminal server for use at the Coachella Valley
Water District. The information about this configuration is available at
http://sylmar.gps.caltech.edu/stan/net/cvwd.html
- One of the disks on the Spring RAID failed. Replaced it with a cold
spare and contacted nStor [formerly Andataco] about obtaining a replacement.
The failed drive was returned to nStor under RMA# 19326
- Did more testing of the squid server. It was able to put out 48Mb/s in
testing, which is an order of magnitude higher than the peak data rate
after Hector Mine.
- Moved the Squid server into a space in the rack next to Ehzsouth.
- Got the DNS entry for www-socal.wr.usgs.gov changed to point to the
Squid server.
- Moved the plugs for the racks in the Yellow House computer room to the
new generator circuits so that the web servers will be on backup power.
- Put TCP wrappers on terra10 and air.
- Rebuilt the /export/home11 filesystem on gobi. Needed to do 'newfs -i
512' to build it with more inodes.
- There was a DNS problem with the ehz subdomain servers on Thanksgiving
Day. Discovered that our Menlo Park phone list is out of date. Was
finally able to reach Stan Silverman, who was able to correct the problem
by issuing a HUP command to each nameserver.
***
- Upgraded VMS on the primary online system.
- Set up a Squid reverse-proxy caching server to act as an amplifier for
the main office web server. In testing, it appears to increase performance
by about a factor of ten. Put it into production on Tuesday, November 23.