I am in the process of replacing Eskimo’s mail servers. I’ve got one new server up and running for inbound, working on the 2nd, and then the client mail server will be next.
It’s a bit messy and that is pretty much unavoidable. The new servers are i7-2600 based, running CentOS 6.4, and completely different software. They are using Postfix instead of Sendmail, and are using clam-av and spam-assassin. They use Procmail as the delivery agent, hence it will no longer be necessary to have a .forward to a procmail_wrapper; everything will automatically direct through Procmail for delivery.
When everything is done, you will be able to train spam-assassin by mailing examples of good mail and bad mail to addresses for that purpose or by pressing a spam button in webmail. You will be able to set individually the threshold at which mail is considered spam and specify the handling of it via procmail rules.
Viruses are rejected outright but a mail is sent back to the sender telling them there mail was rejected and which virus it contained so that they can take the necessary actions to disinfect their computer.
I urge your patience while this process takes place; in the end it will be worth it, far less spam, less false positives, and more control over your mail, and the client server will have proper certificates so picky clients will work properly. But getting there is somewhat messy. It will probably take 2-3 weeks to complete this transition.
Because the software is completely new, this is a learning curve for me as well. There are many components and unanticipated interactions can occur.
Due to failure of the old SS-10 running RedHat 4.2, UUCP infrastructure has been upgraded to a virtual machine running on an I7-2600 platform. The new OS is CentOS 6.4 Linux. The UUCP is now Taylor UUCP version 1.07, and the MTA is Postfix.
I haven’t been selling UUCP to new customers in need of that service because I was afraid that 20-year old box was going to die soon and I had not been successful at getting UUCP to work properly with modern Sendmail. Using Postfix as an alternative provided a solution.
We will now be offering UUCP connectivity to new customers, either by local dial-up or over TCP at an annual rate of $264.
If you are interested, please e-mail email@example.com or give us a call at 1-206-812-0051 or Toll Free 1-800-246-6874.
Comcast has taken to blocking port 25, the normal outgoing SMTP port. You can work around the problem by changing your outgoing SMTP port either to port 2525 or port 587.
I would also encourage you to let Comcast know this is unacceptable. Otherwise, they may continue to block additional ports and your Internet functionality will continue to degrade.
The maintenance activity for tonight has been completed. The newer Intel servers including shellx and the web server, the name servers, and the main file server are all updated to Centos 6.4 Linux.
Our web server has been upgraded to Apache 2.4.4. This is mainly a bug and security fix. It fixes several instances that can cause Apache to core dump as well as some cross-scripting exploits.
Tonight’s maintenance activity which involved a huge number of software upgrades, as many as 645 on some servers, and kernel upgrades on all the modern machines, has finished.
The telephone numbers have been ported from Integra to Comcast. There should be no further interrupt….
This mornings maintenance activities have been completed. Primarily kernels were updated on a number of machines and a number of non-kernel updates applied as well.
I rebooted mail and ftp/www today at 1:15 PM Pacific time because it was the quickest way to stop a virus from propagating.
Last night I went to CraigsList looking for a cheap streaming device, preferably a Wii console, and I’m not sure if I went to the right place (I went to www.craigslist.com). I got a banner telling me I’d won an Ipad and then it sent me to a bunch of surveys which were virus laden and managed to infect the Mac workstation I use with four viruses, (this is the first time I’ve seen viruses successfully infect a Unix based system since 1995). It actually infected some aspect of TenFourFox, a PPC power of Firefox web browser. If I shut the browser down the virus is inactive.
Sophos anti-virus detected it, but at least when run from an ordinary user account it wasn’t able to delete it. I’m running a full-scan now from the System Admin account with more aggressive settings on the scan hoping that will remove it, if not I’ll have to go virus hunting manually.
At any rate, if you go to CraigsList and get a screen telling you that you’ve won an Ipad, DON’T FOLLOW THE LINK, close the tab and start over.
If you find your mail session in Pine or an IMAP client suddenly goes read-only, or if you see the “delete” button disappear on web mail, that is a symptom of this virus because it causes your browser to open an IMAP session with your account information and that interferes with your real session.
And actually there were two copies of MyDOOM and two other viruses and I don’t at this point know which was doing what. But whatever these are, they exploit something in Firefox and can affect Mac’s (and I’d be willing to bet by extension Linux since the protections and privileges of the two operating systems are pretty much the same).
Eskimo North will be undergoing some necessarily disruptive maintenance early Saturday February 16th, shortly after midnight. I will be upgrading Linux kernels on a number of machines which requires a reboot on the upgraded machine. Additionally, mail, which has NFS dependencies on the upgraded machine will need to be rebooted.
The reboots should only entail a brief outage of several minutes, however, the mail server will need about ten minutes (it’s an older machine and work is in progress to replace it) but occasionally it does not go down smoothly and requires an in person kick in which case that will add about half an hour to the mail downtime.
The machines being rebooted that will have service impact are Iglulik which serves files for all the other machines, ftp/www, shellx, and ultra7/mail.
If you are on one of the other shell machines and shortly after midnight your session freezes, if you just wait it out it should come back within about five minutes.
Shellx will first experience a freeze like the other shell machines while Iglulik reboots, then it will be rebooted and you will need to re-login.
After a boot, nx may require several attempts to connect successfully. It seems to come alive in stages and takes longer than the rest of the server to stabilize.