serverfault.com

I'm unable to receive any emails on my email server at jessiekirk.xyz. I setup Postfix and Dovecot on my Raspberry Pi 4 (Ubuntu). I followed along with the tutorial at samhobbs.co.uk, deviating on Dovecot setup as its been updated since the tutorial was written. I was able to receive emails by SMTP and SMTPS, send via SMTPS and check via IMAPS (using FairEmail). Next, following instructions on EasyDMARC, I set up DMARC, SPF and DKIM (via OpenDKIM). I also added an MX record and added email aliases to comply with RFC2142. I'm now not able to receive emails, though I can still send them. Things I've tried to fix it: Checked

I came recently accross one bug which after 3 day venture I have not figured out yet. I run mysql on a Ubuntu 24.04 LTS base server. The process crashed and rebooted whole system due to insufficient memory. It happened already in the past many times. It just reboots and then it runs again. However this time the server has crashed and since then it became unpingable. I contancted my provider and they have told me I have issue in GRUB. I went to rescue mode, mounted all disks and started inspecting. I reinstalled GRUB, upgraded etc. but none of it worked. After struggling I reinstalled kernel and the server became pingable. Even ssh worked. After that I have rebooted the server and I tried ssh into it. But I got "connection refused error". I inspected the ssh, firewall, iptables... All seemed ok. Server was up, network was up, just the port 22 was closed. After another reboot, the server went unpingable. After struggling for day I gave up and installed l

I am looking for a way to temporarily disable or stop an Azure Cognitive Services resource to ensure no further requests are processed and to halt consumption-based billing. How can one disable/pause an Azure Cognitive Services resource to prevent further billing on that resource?

¿Cómo optimizar la configuración de memoria RAM y uso de swap en un equipo de producción que ejecuta un servidor web Apache y una instancia de MySQL? Actualmente tiene 16 GB de RAM, pero se observan picos de uso que causan lentitud en las operaciones, y quiero evitar que el sistema use excesivamente el swap.

I am setting up an Azure OpenAI resource and I need to restrict access so that only requests from our office static IPs and our application server are allowed. By default, the resource seems to be open to "All networks." How can I configure an Azure OpenAI resource so that it allows requests from certain IPs? E.g., openai.com supports allowlists. From https://platform.openai.com/settings/organization/security/ip-allowlist:

In my enterprise, we have Linux hosts with SCSI HBAs connected to SAN via Fibre Channel ('qla2xxx' driver). Fibre Channel uses switched fabric topology. Sometimes, there are new LUNs provisioned to our hosts from the SAN side, and sometimes those LUNs are not discovered, so we need to scan HBAs: nothing special. I'm not an expert neither in SAN, SCSI, or Fibre Channel, so I've read through some documents and specifications. As I understand it, Fibre Channel has three basic topologies: point-to-point, arbitrated loop, switched fabric. Now, to discover new LUNs via the FC interface, there is a possibility to use the 'issue_lip' file, and there is also the corresponding option for the 'rescan-scsi-bus.sh' (it just writes into the same file). As all Fibre Channel specifications are hidden under paywall, I cannot check what's the reason to use LIP when we have FC with the switched fabric. LIP stands for Loop Initialization Procedure (or process, or primitive: whatever), so if we

I have two hard drives of the same model: HGST HUH721010AL5200 (10TB SAS). Both are declared as 10TB by the manufacturer, but the reported usable sizes differ: Drive A: 9.095 TB visible Drive B: 8.909 TB visible My server requires at least 9TB per disk, so the smaller one will not work. Images: Disk 9TB+ Disk 9TB+ stats Disk 8.9TB Disk 8.9TB stats I was following these instructions: https://talesinit.blogspot.com/2015/11/formatted-with-ty

I want to make a valid TLS certificate for my MX server but I'm confused which hostname/SN I have to set. The MX dns record is published with a cname to serve multiple domains. Example: MX mail.mydomain.com CNAME mail.mydomain.com > server1.otherdomain.com A server1.otherdomain.com x.x.x.x So I don't know which hostname I have to use for the tls certificate. Should I use the mx name? (in that case it is different by domain), or maybe the match is checked using the ptr resolution? Or the name is smtp Helo exchange?

I have certbot that keep a valid certificate and key in its folder /etc/letsencrypt/archive/ The cert is public (644) and key is restricted to root (600). I'm wondering how can I let exim read as TLS certificate. Exim doesn't run as root and cannot read the key. I imagine is not a good idea to change key permission to a group with exim use, what is the best practice? Does certbot like to change its key permission? What happen at key renewal?

I’m dealing with a customer which uses a proxy for Internet access; an explicit proxy, not a transparent one. You need to actually call it, and the configuration is deployed to all users using a proxy.pac file and Group Policies. Now, the customer wants to use some Azure services with Private Endpoints (I’m specifically interested in services using HTTPS connections). The networking and DNS resolution are in place, but the proxy gets in the way, because it can’t access those Private Endpoints (they don’t want everyone to be able to connect to them, thus they are filtering access; allowing the proxy would allow all users going through it); this means all actual users of those services need to bypass the proxy when connecting to them. This is easily (enough) achieved for a web browser by setting proxy exclusions, either in the proxy settings of the client or in the proxy.pac file. It’s cumberstome, but it works. But here’s the