- QDDS was hung on Rift. The console showed lots of 'out of memory' errors.
Killed it and restarted.
- Ned reported a problem with subscriptions to the RELM mailing lists. This
turned out to be caused by the redirect for 'mailing_lists.html' in the
Apache configuration on Agent86. It was redirecting the CGI call to the
main server copy of ml-sub-requests, which was sending the subscription
requests to the wrong address. The brute-force solution was to change the
name of the html file under the RELM directory to avoid this conflict.
- Rebuilt the old Program web site Squid server as a development server
for the Program and Pasadena web sites.
- Put the rewritten remove_old_datafiles.pl script on Jet. This improved
the speed of cleaning out old files in the wavepool.
- Post-mortem meeting on the October 30 network problems.
- Moved Pluton to the telemetry room and changed its IP address to 65.201.
- Meeting to discuss Trinet intranet web server issues.
- Got the replacement power supply for the nStor RAID on Spring.
- Increased the number of nfsd processes on Pluton from 4 to 16. This
improved performance somewhat. It was done by adding the following line
to /etc/rc.conf:
nfs_server_enable="YES"
nfs_server_flags="-u -t -n 16"
- Holiday an Monday.
- QDDS was hung on Fang. Set up a cron job to monitor the QDDS log file.
The job runs every five minutes, and if the log file has not been updated
since the last cron run, it kills and restarts QDDS.
- Upgraded Bort to FreeBSD 4.4.
- Helped Stan Silverman with a patch for ssh 1.2.27.
- Called Akamai to ask about possible service for trinet.org.
- Called Akamai to ask about slow web service on quake.wr.usgs.gov when
ehzeast was down.
- Set up a scign account and scignmail mailing list on eqinfo.
- Got the disk from the Reston Program Squid server. Rebuilt it as
'horst.wr.usgs.gov' and set it up to be a web server for the Program
and Pasadena sites.
- Added the rest of the Pasadena office personnel to the eqalarm-local
mailing list.
- Talked to Khasmir at Akamai about slow service when the Reston
web server was down. She said that the default connection timeout
is 2 minutes, and the service timeout is 10 seconds. Asked her to
change the connection timeout to 10 seconds.
- Got the disk from the Menlo Program Squid server. Rebuilt is to
be 'graben.wr.usgs.gov' and sent it to Reston to be a web server for
the Program and Pasadena sites.
- Had the MX record for scign.org changed to point to eqinfo.gps.caltech.edu.
- Put an updated version of CNSSM on Bort. Had to hack a few minor
things to port it to FreeBSD. Details are at
http://bort.gps.caltech.edu/stan/mail-archive/msg00037.html
- Researched possible upgrades for Jet and Spring. We could add
memory to the systems to reduce swap usage. CPU upgrades depend on
the clock board and whether it can handle faster processors.
- Populated the scignmail@scign.org mailing list and got the archives
from JPL. The archives are now stored at
http://www.scign.org/mail-archive/
- Thanksgiving holiday Thursday and Friday.
- Stan Silverman set up the former Squid server in Menlo Park with
the disk I had configured and sent out last week. The new machine
name is 'horst.wr.usgs.gov', and it will take over as one of the
origin servers for earthquake.usgs.gov.
- Heard back from Khasmir at Akamai about EdgeSuite. She set the
connect timeout down to 10 seconds for all our sites. Also, she
said that the Akamai servers will round-robin requests between
multiple origin servers.
- Upgraded the OS on eqinfo to FreeBSD 4.4. This is the mail server
that handles the relm.org and scign.org mailing lists, as well as
functioning as a backup server for the earthquake mailing lists.
- Upgraded sshd on several machines. There was a minor glitch in this,
as I had to remove the file '/usr/local/etc/rc.d/sshd.sh' and add the
line 'sshd_enable="YES"' in /etc/rc.conf. Also, it was necessary to
add some lines to /etc/pam.conf to make it work.
- Added two new 18GB disks to Iron. These will replace the old 18GB
disks that hold the /home and /opt filesystems.
- Set up a script to copy CIIM input files over the network to Genie for
processing. This is so that Genie does not have to NFS mount the disk
from the server. This will allow us to put servers in remote locations.
- Judy Konnert brought up the former Squid server in Reston with the
disk I had sent to her last week. The new machine name is
'graben.er.usgs.gov' and it will become the new origin server for
earthquake.usgs.gov.
- Removed Ojai from the FDDI ring.
- Fixed a typo in the scignmail archiving script.
- Set up a memory-based filesystem for the Trinet Squid server to use
for caching files. This may improve performance.
***
- Rebuilt the three Squid servers that were used for Earthquake Hazards
Program web site. One is now set up as the development server for
the Pasadena and Program sites. The other two are set up to take over
as the main servers for the Program site, replacing the old Suns that
are currently doing this job.
- Set up the eqinfo mailing list server to act as a mail server for
scign.org. Set up a mailing list on it to take over for the old
SCIGN mailing list at JPL.