Sender | Message | Time |
---|---|---|
25 Apr 2024 | ||
Tim Jones | No idea, sorry, better to ask the OpenEBS people | 21:07:18 |
aeonremnant | Has anyone used Mayastor with RDMA on Talos? Looking for performance characteristics and it's been hard tracking any down. | 21:12:18 |
Robert Rotter | Damn, bad advertisedSubnets: - 192.168.45.0/24 <------------------ in etcd | 22:49:24 |
Robert Rotter | Yuhu, running local and cloud controlplanes together via Wireguard | 22:54:02 |
Dan Korel joined the room. | 23:19:36 | |
Dan Korel set a profile picture. | 23:19:38 | |
aeonremnant | * I appreciate the help mate! Just setting up a zone for my DNS and a simple A record was enough for it! Just incase people need it:
After this the Talos installs should automagically discover Omni and assign like normal, just VERY fast. | 23:24:36 |
Sina Karimi joined the room. | 23:56:43 | |
Sina Karimi set a profile picture. | 23:56:44 | |
26 Apr 2024 | ||
Dan Korel changed their profile picture. | 01:05:21 | |
Tim Jones | In reply to@_slack_taloscommunity_UG8G8UMMG:matrix.org(edited) ... e.g. ```10.0.0.22: {"ts":1714075908009.4011,"caller":"kubelet/kubelet_node_status.go:92","msg":"Unable to register node with API server","node":{"name":"talos-hybrid-worker-0"},"err":"Post \"<https://127.0.0.1:7445/api/v1/nodes>\": tls: failed to verify certificate: x509: certificate is valid for 10.96.0.1, 13.41.56.39, 14.0.149.230, fd5c:6e1e:2983:a702:8ec:3fff:fefb:d915, not ... => ... e.g. ```xxx: {"ts":1714075908009.4011,"caller":"kubelet/kubelet_node_status.go:92","msg":"Unable to register node with API server","node":{"name":"talos-hybrid-worker-0"},"err":"Post \"<https://127.0.0.1:7445/api/v1/nodes>\": tls: failed to verify certificate: x509: certificate is valid for xxx, xxx, xxx, xxx, not ... | 08:22:45 |
Ram | I did some digging and I see pvc and pv being displayed, and path exists until /var/openebs/ . but not local/pvc-0f3db53f-25e1-4a76-ae5d-2c2e81498771 . however i do see pv and pvc pvc-0f3db53f-25e1-4a76-ae5d-2c2e81498771 listed when i get and they are bound . could this be something to do with talos? 🤔 | 10:19:40 |
Tim Jones | I don't know enough about how OpenEBS manages the volumes on the file system to answer. As far as Talos goes, if you're just mounting a user disk, make sure it's the same disk (i.e. using /dev/disk/by-id ) | 10:22:27 |
Ram | > if you're just mounting a user disk, make sure it's the same disk
sorry, if this is referred to having mention extraMount in machineconfig, its included already.
extraMounts: - destination: /var/openebs type: bind source: /var/openebs options: - bind - rshared - rw | 10:24:54 |
Tim Jones | You're just using the system disk as backing for OpenEBS? | 10:25:31 |
Ram | yes | 10:25:57 |
Ram | i was under an impression that if we use --preserve=true flag, the data would stay intact. | 10:27:15 |
Tim Jones | that's really not a good idea. Talos should "own" the entire system disk, and you should assigned secondary storage user disks for workloads. Many operations in Talos will just wipe the system disk as it's assumed to be ephemeral. | 10:27:21 |
Tim Jones | > i was under an impression that if we use --preserve=true flag, the data would stay intact. it will, but that doesn't protect you from forgetting that flag on a command and wiping the disk entirely | 10:28:21 |
Ram | > you should assigned secondary storage user disks for workloads can i ask if you can provide bit more on this, please? sorry, not following. does secondary storage mean taking backup using velero or similar and keep it in object stores like s3? | 10:33:09 |
Tim Jones | no, just another block device/disk to the machine | 10:33:28 |
Ram | just to clarify. is it this? | 10:35:28 |
Tim Jones | yes | 10:36:02 |
Ram | ok its there then. sorry i misunderstood earlier | 10:36:17 |
Injo | I am facing a network issue with (only) Talos nodes. I recently got a new switch, Aruba 1960 and the 5 Talos nodes see regular links going down and back up. Im looking at the machineconfig documentation but I don't see whether its possible to force 1Gbps fixed, instead of the default autosensing. Is this possible? | 12:29:15 |
Serve | Is it correct that the virtual IP only moves to another machines when executing
talosctl reboot | 12:42:24 |
Tim Jones | Not really, I normally would force this on the switch side than the host... | 13:27:54 |
Tim Jones | not only, but it's better to allow Talos warning that it is going to shut down so it can do so cleanly than "pull the plug" so-to-speak. Basically the VIP is assigned via etcd election, so cleanly shutting down the node allows Talos to denounce leadership of the VIP and initiate an election and hand over the VIP with minimal interference. If you just kill the Talos machine with the VIP, there will be a noticeable delay in that process since the leader election has to time out first before a new Talos VIP leader can be elected and assigned the VIP, but it will happen eventually. | 13:31:19 |
Bianca Pereira | Yes, that'd be great! | 14:38:09 |
Sean Saperstein | How about Monday 28th? We have time midday (10am Pacific). Would that work for you? | 14:51:31 |