In GitLab by @paul.laffitte on Dec 28, 2020, 18:58

@abuisine: il faudrait pouvoir ajouter un nodeSelector dans le helm chart pour limiter le daemonset aux noeuds d’une availability zone par exemple

In GitLab by @paul.laffitte on Dec 29, 2020, 18:20

• ~~Add levels of verbosity~~
• ~~Logs in Dothill at lower levels of verbosity~~
• [x] Log output of multipath -f on error (map in use)

In GitLab by @arcln on May 10, 2019, 22:05

• [x] read parameters from storage class
• [x] implement lazy login/logout
• [x] fetch storage class from its name in Delete() call

In GitLab by @paul.laffitte on Sep 21, 2020, 23:16

• [x] The file /etc/iscis/initiatorname.iscsi may not exists and cannot be created by the script without sudo
• [x] The script should crash if environment variables for secret templating are not available
• [x] Delete sockets in /tmp after a crash

In GitLab by @paul.laffitte on Sep 15, 2020, 23:01

Since klog seems to remove \n at the end of a line, if a user encode his username with a trailing \n (echo "username" | base64), he will not notice it in the logs. Worse, it can confirm a fake belief that his username is correctly formatted.

The goal is to help users to find unwanted characters in their username (, \t or \n for instance).

In GitLab by @paul.laffitte on Nov 4, 2020, 18:38

https://github.com/container-storage-interface/spec/blob/master/spec.md#volume-lifecycle

• ControllerUnpublish is executed after NodeUnpublish
• ControllerUnpublish unmap the volume on the MSA
• NodeUnpublish rescan after ejecting devices, so they reappears just before being unmapped

Hi there !

I am using Dell Powervault ME4024 disk storage which is claimed to be supported, I try to use the driver but I encounter an authentication problem to the API.

Controller logs :

I1103 14:36:37.346453 1 driver.go:112] === [ROUTINE START] /csi.v1.Controller/CreateVolume ===
I1103 14:36:37.346501 1 controller.go:243] using dothill API at address https://sanip/api
I1103 14:36:37.346513 1 controller.go:245] dothill client is already configured for this API, skipping login
I1103 14:36:37.346528 1 provisioner.go:67] received 5368709120B volume request
I1103 14:36:37.346539 1 provisioner.go:76] creating volume fde4b27e9ae44ddd9750fd92c4368b17 (size 5368709120B) in pool A
I1103 14:36:37.346568 1 dothill.go:94] -> GET /login/<hidden>
I1103 14:36:37.719588 1 dothill.go:124] <- [2 Error] <hidden>
E1103 14:36:37.719723 1 driver.go:118] Dothill API returned non-zero code 2 (Invalid sessionkey)
I1103 14:36:37.719737 1 driver.go:121] === [ROUTINE END] /csi.v1.Controller/CreateVolume === 

Kubernetes version : 1.24.4

API documentation : https://www.dell.com/support/manuals/fr-fr/powervault-me4024/me4_series_cli_pub/using-a-script-to-access-the-cli?guid=guid-9ae5ccd6-a207-42df-b2f3-1e02a487a354&lang=en-us

Can you help me?

Thanks !

Would it be possible to provide the image

enix/san-iscsi-csi

as a multi arch image covering

• amd64 (as now)
• arm64 (additionally)

Background: I have a hybrid cluster with amd64 and arm64 nodes and the daemonset does not run on arm64 nodes.

Good day!

I am using HPE MSA 2060 disk storage which is claimed to be supported.

I took the HPE MSA 2060 API documentation here - https://www.intesiscon.com/ficheros/manuales-tecnicos/255-HPE-a00105313en-us-HPE-MSA-1060-2060-2062-CLI-Reference-Guide.pdf

Faced some problems while using this project.

1.https://github.com/enix/san-iscsi-csi/blob/main/pkg/controller/publisher.go#L43 does not match the HPE MSA 2060 API documentation. Here, instead of host-maps should bemaps(347 documentation page) 2. If you connect several PVs with one Helm deployment, then all volumes in the disk storage receive the same LUN number, which is unacceptable. 3. For some reason, the partition table and file system are not created in the slave PV. 4.https://github.com/enix/dothill-api-go/issues/12

Applications used:

1. Ubuntu 20.04
2. Kubernetes 1.21.4
3. Helm 3.3.4

PS: If you need additional information - indicate what you need to provide

With structured logging, log messages are no longer formatted, leaving argument marshalling up to the logging client implementation. This allows messages to be a static description of event.

All string formatting (%d, %v, %w, %s) should be removed and log message string simplified. Describing arguments in log messages is no longer needed and should be removed leaving only a description of what happened.

https://github.com/kubernetes/community/blob/master/contributors/devel/sig-instrumentation/migration-to-structured-logging.md https://kubernetes.io/blog/2020/09/04/kubernetes-1-19-introducing-structured-logs/

While running 2.4.1 and expanding volumes, I get a warning event on the PVC ressource :

• reason : ExternalExpanding
• from : volume_expand
• message : Ignoring the PVC: didn't find a plugin capable of expanding the volume; waiting for an external controller to process this PVC

after a rather short (30s) period of time, the resize is being processed as it should