The problem isn’t limited to the redhat machines, it seems EVERYONE pushed out a new version of rdp without bothering to test and the new version expects a newer version of openssl than that which is supplied. So to get it working again I’m going to have to recompile openssl on all the machines. May be down for a while.
Category Archives: Uncategorized
Last Night’s Maintenance Mint and Mail
I fell asleep while still working on system maintenance which resulted in some problems with sending mail, incoming mail, and mint. I apologize, went past what caffeine could do.
rdp / vnc on Centos7/8/Stream/Fedora Unavailable
RDP and VNC (which depends upon RDP) is unavailable on Centos7, Centos8, CentosStream, and Fedora.
The reason for this is that they sent an upgrade to the RDP package that was built against openssl that is newer that the version that exists in the repository so it fails.
Maintenance Work Tonight
Later this evening, possibly after midnight, I will be taking a number of servers down for a short time to audit packages.
One of our customers private virtual server crashed today and after rebooting was consuming excessive CPU. An audit of the package system showed there were many missing and obsolete packages even though it had been updated regularly. After fixing this CPU usage returned to normal.
This prompted me to look at two other private virtual servers that had high CPU usage and I found the same corruption. And again fixing it, brought CPU utilization down to normal levels.
Because most of the system servers are also based upon Ubuntu 20.04, the same operating system that these were running, and because I loaded the system machines from the same ISO, I am going to take various ubuntu based system servers down to check for similar corruption. This check only takes about five minutes per server if there is nothing wrong and up to half an hour if there is, but NFS and NIS relationships may also take some time to restore afterwards.
Maintenance Completed
Tonight’s maintenance is complete.
Reboot Wednesday 11pm PST
I am planning a server reboot on Wednesday January 27th at 11pm. I expect all services will be restored by 11:30pm.
This is for a kernel upgrade from 5.10.4 to 5.10.10. It fixes a number of bugs that are minor and thus far haven’t affected us and probably won’t but prefer to have infrastructure be as clean as it can.
pam-abl used for DoS attack.
Pam_abl or libpam-abl is a pam module used to black list users or IP addresses that repeatedly fail authentication. It serves the same function as fail2ban but has a flaw that it can be used to deny service to a user by repeatedly trying passwords for that user from different IP addresses. It was used to deny a number of customers from using webmail and in the process of troubleshooting I also broke the mount point for mail used by ubuntu, so both webmail and ubuntu were inaccessible for a while. This module has been disabled since it can be abused this way and is redundant as it serves the same function as fail2ban.
Mail and other Login Difficulties
We are being hit with such heavy brute force password guessing attempts that it has triggered pam_abl module on a number of hosts, a situation I have not previously encountered, and I am trying to figure out how to reset it and set it high enough that fail2ban will trigger first.
Eskimo North’s History
New Kernels
I’ve compiled and placed online the latest mainstream Linux kernels 5.11rc4 and 5.10.9. The 5.11rc4 is only available in client form since it is still a release candidate it is not really suitable for server use yet. When the official release comes out I will place a server version.
The kernels can be obtained from: https://www.eskimo.com/kernel/linux-{version}-tickless/{client|server}/files.deb.
There are three files for each kernel, download them all and install with: dpkg -i *.deb These will work on any debian based distribution such as Debian, Ubuntu, Mint, MxLinux, Zorin, Julinux, etc.
These are all compiled as fully tickless kernels. This is advantageous in situations where not wasting CPU cycles servicing clock interrupts where there is no work to be done is advantages, such as laptops where battery life is a concern or hosts with many virtual machines where each virtual machine wasting CPU totally can end up wasting more than real applications.
The “client” kernels are fully pre-emptive with a 1000Hz clock rate to minimize latency and are the best choice for multi-media, gaming, home desktops and workstations. The “server” kernel is non-pre-emptive with a 100Hz clock to maximize throughput at the expense of latency.
These kernels are all capable of running on bare metal or virtualized environments. They are all capable of hosting KVM/Qemu virtual machines provided you have the proper vt-t/vt-d instructions in your CPU.