2018

September 2018

September 9, 2018:

Previously, I had a problem with nextcloud where it wasn't able to be recognized. I went to a backup and it turned out that this was on the nextcloud side. It had nothing to do with my nginx reverse proxy which was where I assumed the issue initially was.

I went ahead and kept on the backup and then synced the local datastores on my laptop and desktop to bring the data up to date on the vm even if the database records were going to have a month of gap.However, my sharelatex docker container was also running on that vm. Unfortunately, I had done quite a bit of work in my sharelatex docker so I needed to get that back. The solution I adopted was to make a backup then revert to the backup I made before the above stuff went awry and then boot the container. After downloading the zipped project folders for all my projects, I shut down that vm again and restored to the first backup I made. I then uploaded the project folders to the sharelatex docker container again. I would like to get away from docker as it cannot be run in an lxc container easily. When I do run it I am getting errors like `oci runtime create failed debian`. This thread displays some of the modifications that can be made to proxmox's lxc config files to fix this. I modified the config file of my test container to have all but the last line. I suppose I should try with the last line and see what happens.

In addition, I modified the max number of backups on the 1tb western digital drive I have in its own raid 0. Containers in Proxmox will replace the older containers if the number of backups after the current backup job will exceed what you've configured. VMs do the opposite. which is dumb. They won't back-up any more once you've reached the limit. I'll have to remember to clean out old backups for my single vm.

September 20 2018

I tried installing a matrix homeserver. However, this didn't go great. Will try again this weekend.

I restarted Mwenyeji today. I noticed that the fs.inotify changes I had made were no longer in affect. I started getting file handle errors again. To fix this, I ran the following.

root@mwenyeji:~# echo "fs.inotify.max_queued_events=48000"|sudo tee -a /etc/sysctl.conf
root@mwenyeji:~# echo "fs.inotify.max_user_instances=512"|sudotee -a /etc/sysctl.conf
root@mwenyeji:~# echo "fs.inotify.max_user_watches=120000"|sudo tee -a /etc/sysctl.conf

I setup a cronjob to take a snapshot of the wiki every night since I cannot figure out if there is any version controlling going on behind the scenes in bookstack. snapshots of the wiki appear to be very small so they don't take up much space having a lot of them.

November 2018

ucs c240 m3

I purchased a cisco ucs c240 m3 on ebay for $195. The unit came equipped with dual e5-2609's, 8gb of ram, heatsinks and four drive trays. Initially, the fans were too loud so I upgraded the firmware using the host upgrade utility.

This allowed me to bring the fans down to a very reasonable level. However, after running linpack on the machine, I noticed that it was rebooting after being under heavy load. I checked out the ram and everything but the same thing happened. None of the capacitors on the motherboard appear swolen or leaky.

I tried replacing the processors (since the unit wasn't enclosed in electrostatic wrap during shipping I figured that the processors may have experienced some static discharge rendering them unreliable.

I was getting errors in cimc like PVCCP_P1: Processor 1 voltage is lower critical before I changed the procs. I am still getting reboots after the server has been under heavy load. It's now not while the processors are stressed but actually after the process has been killed and the machine idles for a while.

Running geekbench is fine, running linpack is too much. I think I have no choice but to return this server, which is a bummer because I'm going to be out my shipping costs.

Update

Upgrading the processor energy profile to high performance has prevented failures. The longest I've been up has been 8 hours. I'm quite confident it can stay running for longer now. I wanted to work up to something longer since a system failure results in the fans hitting 100% which can be quite annoying in the middle of the night.

December update

Unfortunately, the problems resurfaced after attempting to fill all the ram slots on processor 0. Stepping back down to only 4 dimms fixed the issue. I still suspect an issue with the powersupply. I have gotten $60 from the seller due to the situation. I hope to buy a new psu for this machine using this money.

October 2018

Quanta windmill board 2

I got another board but it appears to not be working that well. I also got a noctua fan controller which is working great. However, after writing the bios chip using the ch341a chip I bought off ebay, the sata connectors are not working. I am able to boot an operating system fine from the usb port, I just cannot access the sata connectors well.

The next course of action is updating the bios using the freedos updater. If the sata connection problem is not fixed after that, I may have to buy a pcie raid card to avoid using the onboard sata controller.

December 10

Unfortunately, the situation only deteriorated with the bios chip. It lost all ability to post. I am afraid I will have to buy another windmill board from the seller.