serverfault.com

We have a bunch of Oracle Linux 9 machines, many of which have MariaDB installed. It was originally installed via dnf (using ansible) - name: Install mariadb (mysql) dnf: name: - "mariadb-server-10.5*" - perl-DBD-MySQL state: latest Our deployments always end by updating everything via dnf - name: upgrade all packages dnf: name: '*' state: latest This morning this started failed. Running dnf update manually also failed. We had some updates run in the night OK but a bit after 0600 UTC this morning it started failing. Error is Error: Problem: package mysql-libs-8.0.43-1.el9_6.x86_64 from @System requires mysql-common(x86-64) = 8.0.43-1.el9_6, but none of the providers can be installed - cannot install both mysql-common-8.0.44-1.el9_7.x86_64 fro

I have a k3s cluster running Traefik and Cert-Manager to handle TLS termination. In front of k3s I run haproxy on PFSense as a load balancer. HAProxy uses proxy protocol to preserve source IP. This allows me to automate the process of bringing up a new service using only k8s constructs, and I never need to manipulate HAProxy for each site. I'd like to support HTTP3/QUIC. I know that HAProxy's free edition doesn't support it, and I think that even if it did it would have to manage the TLS termination itself. My question is, is it even feasible to do this with any technology, given how QUIC works? With TCP it's easy to separate the connection from the content of the stream, enabling a smart proxy use something like proxy protocol to add context like source IP. But in QUIC is there any equivalent, where one could listen on 443 UDP to determine a connection?

I posted this to Bitnami Github help but so far no response and so I am trying here if someone could help. https://github.com/bitnami/vms/issues/1877 Environment: Linux 6.1.0-39-amd64 Debian 6.1.148-1 (bitnami vm), Redmine version 4.2.1.stable, Ruby version 2.6.7-p197 (2021-04-05) [x86_64-linux], Rails version 5.2.5, Database adapter Mysql2 When I try to remove a plugin from this instance of Redmine using the command sudo bundle exec rake redmine:plugins NAME=redmine_agile version=0 rails_env=production then I get an error Mysql2::Error::ConnectionError: Can't connect to local MySQL server through socket '/tmp/mysql.sock' (2). If I run netstat -anp|grep mysql, then it appears that MySQL is on a different so

I’ve read about models like Robust Video Matting (RVM) and MODNet, which seem to focus on per-frame matting combined with temporal consistency across frames. From what I understand, these models are trained specifically for dense pixel-level prediction and video continuity, which feels quite different from the way large multimodal models (e.g. Gemini or GPT-4V) operate. This made me wonder: What is the current state-of-the-art approach for video background removal / video matting in practice? Why can’t large multimodal models directly perform high-quality video matting, even though they can “understand” images and videos? Is this primarily an architectural limitation (token-based vs pixel-based outputs), a performance issue, or a training data issue? I’m trying to understand where the technical boundary is between specialized video matting models and general-purpose multimodal models.

Fast question. Nerdctl report this container running nerdctl --namespace k8s.io image ls|grep -i debian debian 13122025 2d9f95f25185 8 minutes ago linux/amd64 160.4MB 66.63MB debian latest 0d01188e8dd0 3 days ago linux/amd64 133.9MB 49.29MB My question is..how to see it using standard kubernetes tool? This command return nothing kubectl get pods --all-namespaces -o=custom-columns=NameSpace:.metadata.namespace,NAME:.metadata.name,CONTAINERS:.spec.containers[*].name|grep -i debi I have start the container using this command nerdctl --namespace k8s.io container start debian

I am trying to find a list of files older than x days. An example of one such file is: -rw-r----- 1 user1 user1 480K Dec 12 16:25 a_filename.20251212162502 If today is the 17th Dec, then in theory if I run in the same directory as the file: find . -type f -mtime +2 -print it should find the file, right? but it doesn't. If I run stat on the file it also looks ok: stat a_filename.20251212162502 File: a_filename.20251212162502 Size: 491491 Blocks: 960 IO Block: 4096 regular file Device: 254,2 Inode: 13 Links: 1 Access: (0640/-rw-r-----) Uid: ( 1001/user1) Gid: ( 115/ user1) Access: 2025-12-12 16:25:02.364593519 +0000 Modify: 2025-12-12 16:25:00.127955492 +0000 Change: 2025-12-12 16:25:02.368593490 +0000 Birth: 2025-12-12 16:25:02.364593519 +0000 I'm running Debian 12 if that makes any difference. Is there another way to locate th

Dovecot throws error messages after a recent upgrade from 2.3 to 2.4.1 I migrated the previous configuration to 2.4 standards and ran doveconf -n successfully. Dovecot and Postfix services start without errors. However, I get the following message in the logs when receiving a mail. Dec 14 05:57:53 xxxxxxx.com postfix/pipe[1439]: 3D7241F694: [email protected], relay=dovecot, delay=3.7, delays=3.6/0.02/0/0.07, dsn=5.3.0, status=bounced (Command died with status 89: "/usr/libexec/dovecot/dovecot-lda". Command output: doveconf: Fatal: Error in configuration file /etc/dovecot/conf.d/10-ssl.conf line 5: cert_file: open(/etc/letsencrypt/live/xxxxxxx.com/fullchain.pem) failed: Permission denied ) The fullchain.pem cert file exists. I have checked other similar questions on the subject but the solutions do not apply. Does dovecot 2.4 differ in its requirement for file permissions? This is happening on a machine that

I'm struggling with IBM Verify Identity Access 11 configuration, with something that probably should be super straightforward. I have Federation with MS Entra, which I think, based on the logs, seems to work. I want to force SAML authentication on virtual junction, either by button or (even better) on first access and after successful sign in - start sending IV-user header to target backend. I cannot neither trigger this sign in (I always get the standard forms login.html) nor after triggering the sps//saml20/logininitial URL - sending the header to backend server. I've even failed with posting this question on ibm.community - but i've got information "your post will be reviewed" and no sign of it yet - like it didn't happen...

I experienced a full disk issue probably connected to a DDOS attack, the apache access log file was huge. I know logrotation is triggered by default by file size. My server is an Ubuntu 20.04 with default apache and logrotate configuration. My question is: does logrotate checks the file size just when its processes is scheduled, usually once a day by a daily cron job? If yes, this explains how access log can become huge in a single day.

I followed this tutorial to create a shared folder in my Linux guest machine in Virtual Box that can be accessed from my host machine (Windows 11). I am able to access the samba share from Linux by navigating to my own IP but not in Windows. When navigating to the guest machines IP in File Explorer in Windows (host) I get the error "You possibly spelled it wrong" etc. From what I'm finding while searching it looks like I possibly need to add another network adapter to my VM like host-only, but I'm not clear on how to do this. VirtualBox version 7.2.4 Kali Linux 2025.4 Windows 11