Author: Kalle

Categories
Ceph

A working example of a Ceph RBD pool for VMM

<pool type="rbd">
  <name>RBD</name>
  <source>
    <host name="192.168.7.23" port="6789"/>
    <host name="192.168.7.31" port="6789"/>
    <name>rbd</name>
    <auth type="ceph" username="libvirt">
      <secret uuid="ce81a9d5-e184-43f5-9025-9a062d595fcb"/>
    </auth>
  </source>
</pool>

Host elements are addresses to mons. Auth element is a pointer to a registered key. On this webpage you can find information about how to register a Ceph key. https://docs.ceph.com/en/latest/rbd/libvirt/ The first name element is the name for the libvirt pool. The second name is the name of the Ceph RBD pool. In Connection manager in Virtual Machine Manager you have to create a pool and paste the XML into the textbox on the XML tab.

Categories
Linux

Linux dkms with Fedora is always trouble

Dkms modules often breaks with a new kernel. Fedora is fast at using new kernels. If you do a “dnf update” you could get a new kernel and some dkms module fails to compile.

/var/lib/dkms/veeamsnap/5.0.1.4493/build/snapimage.c: In function ‘snapimage_create’:
/var/lib/dkms/veeamsnap/5.0.1.4493/build/snapimage.c:841:24: error: implicit declaration of function ‘blk_alloc_queue’; did you mean ‘blk_sync_queue’? [-Werror=implicit-function-declaration]
841 | image->queue = blk_alloc_queue(NUMA_NO_NODE);
| ^~~~~~~
| blk_sync_queue
/var/lib/dkms/veeamsnap/5.0.1.4493/build/snapimage.c:841:22: warning: assignment to ‘struct request_queue *’ from ‘int’ makes pointer from integer without a cast [-Wint-conversion]
841 | image->queue = blk_alloc_queue(NUMA_NO_NODE);
| ^

Categories
Tech

“Veeam Backup & Replication Community Edition” is a very good backup software

It is the best backup software i have seen. Sadly i used up the 10 licenses immediately. Is is some kind of desktop application. That is good. Some other have a web application. A desktop application can show more information and update it faster. It is easy to add computers to the inventory. Lots of settings and information. It is easy to see what the software is doing and if there was any error. Backups are fast. You can restore files in different ways. It can backup volumes, files and VMs.

https://www.veeam.com/virtual-machine-backup-solution-free.html

https://www.veeam.com/blog/backup-replication-community-edition-features-description.html

Categories
Ceph

Looks like there will always be “high pg count deviation” on Ceph cluster with mixed HDDs and SSDs

I have given up on fixing the “high pg count deviation” on my Ceph cluster. I have a 4TB HDD and a 240GB SSD on one node. The SSD cant hold as many PGs and the HDD do. Maybe i should try to turn the SSD into a cache for the HDD. It looks complicated when i read about it. It is so easy on Storage Spaces direct.

Update: I have set the SSDs as block database and write ahead log for the HDDs. Now only the HDDs are listed and they all have the same PG count.

Categories
Tech

The systemd haters dont give up

It is Debian without systemd. I don’t understand why anybody would want that. Systemd is good. Starting services with scripts is stupid. The scrips can do anything. It is hard for me to know what the scripts do. It is difficult to know if you have stopped all the things a script have started.

Dear Friends and Software Freedom Lovers, Devuan Developers are delighted to announce the release of Devuan Chimaera 4.0 as the project’s new stable release. This is the result of many months of painstaking work by the Team and detailed testing by the wider Devuan community.

https://www.devuan.org/os/announce/chimaera-release-announce-2021-10-14

Categories
Tech

It took me many hours to find out why Firefox did not accept a certificate

I tried to copy a CA certificate from one CA to another CA. At first it looked i had succeeded. I wanted to issue certificates from both CAs with the same keys. After i installed the certificates i got SEC_ERROR_UNKNOWN_ISSUER every time. I did a search with google about that error. It was simple if i did not have CA root certificate with same name in the subject as i had in issuer on the server certificate i would get that error. I checked that i had added the root CA certificate to the trusted certificates in Firefox. It was there and the subject was correct. After building a debug version of Firefox and running it in the Visual studio debugger i understood how Firefox finds CA certs. It calculates a hash on the subject field and look for a CA cert in the cert database with the same hash. It calculates the hash from everything in subject. Length, string type and oid. When i looked at both certificates i noticed a small difference. One had PrintableString as type and the other had UTF8String for common name. That will make the hash different.

Categories
Network

Scale-out file server is not good for sharing files

I thought the Scale-out fileserver role in Windows server cluster would be good for my file shares. I was wrong. Scale-out file server is for some special cases. It is mostly for Hyper-V and SQL server. If you use it as an ordinary file share it will be slow. Saving files to the cluster shares are much faster now that i have changed them to the older File server.

Categories
Network Tech

I gave up on running a two node Storage spaces direct cluster

It is not impossible to get two node S2D cluster to handle a failover, but it is difficult. The nodes must always check if they in the majority of nodes. If a node believes it is not in the majority it must immediately shutdown cluster resources. As soon as i shutdown one node in two node cluster there is doubt on the remaining node if it belongs to the majority. That is why you need a witness. I used a file share witness. One time when i shutdown a node the other node failed to read the file share witness and the cluster went down. I fixed that by removing the witness from the cluster and added it back. A few days later i shutdown one node and the storage pool on the remaining node went offline. It said something about majority disks missing. At first i did not understand what that meant. After reading about storage pool quorum i understood it. A storage pool must always check if it has a majority of disks connected or else it will shutdown. If there is the same count of drives on both nodes a pool can stay up if it owns the pool resource. I had three HDDs and one SSD on both nodes so i did not know how to fix that. That was when i gave up on a two node cluster. A three node cluster is much easier. If one node disappears there is no doubt about it the two remaining nodes are a majority and if the nodes have equal amount of drives there be no question about if the pool have a majority of drives.

Categories
Tech

I bought my first SMR hard drive

I am only going to use it as backup. I bought the cheapest Seagate. Name of the drive is “Seagate Barracuda 4TB”. I started a backup and it was faster than expected. After a few hours when i checked if the backup was finished it had stopped with an error. The error message said that the drive had been removed. It was still visible in explorer. In the system log i could see that the drive had been removed. Strange because it was now connected. I restarted the backup and a few hours later the backup had stopped because the drive was removed. The backup had stopped nearly at the same place as before. That looked suspicious. I started too google about this Seagate drive. It is a SMR drive with an CMR cache. My guess is that the drive ran out of CMR cache after 400GB and got so slow that the SAS card thought it had been removed or failed. I connected it through USB2 instead. The the backup would finish OK but it was much slower. I will keep this drive but i will never buy a SMR drive again.

Categories
Uncategorized

Some Pfizer contract

LEXO-KONTRATEN-E-PLOTEDownload