Isn’t that what Loki is supposed to do?
Prometheus + Loki + Grafana gets you pretty darn far with a rather low footprint compared to ELK.
3 Likes
I’ve never used QSFP+ before. If you use an SFP+ breakout cable, are the individual SFP+ connections usually configurable in the switch, or is the config for the one QSFP+ port simply propagated to the 4 connections?
I believe they are usually sub interfaces, I have never used it before but the shop next door was messing with them(on dell switches)
“In CRS , It will show qsfp1 up (2,3,4 down), but you can see the qsfp1 port speed is 40G.”
So if you are making it 1 40G then it would just be on qsfp1 and if you are splitting it qsfp1-1,qsfp1-2,qsfp1-3,qsfp1-4
2 Likes
anyone know if while using the trial period for esxi… if you’re allowed the use a vCenter appliance with it.
(not the free license, that you can get that doesnt expire, but the trial/eval period you get before entering a license)
anyone know if the 60 day trial for esxi resets after every reinstall?
I believe it does
Guessing homelab?
1 Like
Yes, the trial state will reset on a fresh reinstall. Keep the VMs on a separate datastore and you just have to go through the rigamarole of reinstalling ESXi every 2 months on your homelab boxes to play around.
You can get an ESXi free key, but this has the limitation of assigning a max of 8 vCPU cores to a single VM, no storage API, and cannot be joined to a vCenter Server. Additionally, it’s only for the latest version, which is 7 and they’ve started locking down installing ESXi onto whatever you want in 7
3 Likes
This is how it’s supposed to work, assuming the switch supports breakout cables. If not, they’ll probably just act as an adapter and I imagine only the first port will actually work.
3 Likes
fine with me… i just want a play ground to test out terraform and ansible modules
1 Like
Does anyone have general advice on how to alert to failures with ansible runs?
My thought was to write a generic “failure_notice” handler, which captures the stderror of the failed task and send it as an attachment to an email.
I have an emailer task already written for something else, and perhaps with some modification I can make it report errors. However, I’m not sure if there is a better way to notify on failure strategy.
if anyone has a strategy they’ve implemented in the past and could give me a high level approach I would be most appreciative.
cotton
2 Likes
you cant scroll up?
Unless I’m misunderstanding, @cotton is probably running the ansible playbook unattended and just wants an email alert if something fails… like if you run it in check mode against a production server.
if running with crontab…
you can set a MAIL_TO var and mail should go there… if postfix/sendmail is setup on the host
Need some space for data on my VMs. NFS mount inside VMs or qcow2/raw file on storage server (TrueNAS)? Hypervisor only has basic storage (240GB mirror)
If the vms need to share the data, NFS, if it’s per vm, then setup a block storage mount (zvol on TrueNAS) and share that via iSCSI.
3 Likes
These are unattended and invocation is event driven. Although there is logging, it’s a very busy component, so it’s a needle in a haystack problem. Therefore, it would be easiest to just alert when there is a problem.
I think I just need to write a generic handler to capture failures and then report the stderror and exit code. If I write it well enough, I should be able attempt to try handle the error and if I can’t alert.
3 Likes
Perhaps the better way is to bubble up the error to the controller which is invoking the playbooks and then have that handle reporting the error. That might be the more appropriate way to handle it. That way I’m not doing error handling in YAML but in (insert favorite imperative language — ie go or whatever).
1 Like
Can you implement a listener (maybe just an fd) that it can dump errors into that simply pipes to sendmail (or whatever).
1 Like
Does that mean just create an “open” fd and pipe stderror to it?
1 Like
Does ansible’s informational output go to stdout or stderr? I think you’ll need some parsing, but yeah either an fd or a socket or something with a separate service that listens, parses and sends alerts.
1 Like