Thanks again. I'll do some speed tests. In my case, I need ready-to-use images (debian, centos, ubuntu, pfsense, kali, windows 10, windows 2016), sometimes big images. This is why I am trying to find out what is the best solution with the use of an iscsi bay. Ah .... if I could change that and use disks and ceph ... Franck
Le 23 nov. 2021 à 09:14, Ignazio Cassano <ignaziocassano@gmail.com> a écrit :
Franck, If the cache works fine , I think glance image format could be qcow2. The volume is created in raw format but the download phase is executed only the fisrt time you create a volume from a new image. With this setup I can create 20-30 instance in a shot and it takes few minutes to complete. I always use general purpose small images and colplete the instance configuration (package installation and so on) with heat or ansible. Ignazio
Il giorno mar 23 nov 2021 alle ore 08:57 Franck VEDEL <franck.vedel@univ-grenoble-alpes.fr <mailto:franck.vedel@univ-grenoble-alpes.fr>> ha scritto: Ignazio, Radoslaw,
thanks to you, I made some modifications and my environment seems to work better (the images are placed on the iiscsi bay on which the volumes are stored). I installed the cache for glance. It works, well I think it does.
My question is: between the different formats (qcow2, raw or other), which is the most efficient if - we create a volume then an instance from the volume - we create an instance from the image - we create an instance without volume - we create a snapshot then an instance from the snapshot
Franck
Le 19 nov. 2021 à 14:50, Ignazio Cassano <ignaziocassano@gmail.com <mailto:ignaziocassano@gmail.com>> a écrit :
Franck, this help you a lot. Thanks Radoslaw Ignazio
Il giorno ven 19 nov 2021 alle ore 12:03 Radosław Piliszek <radoslaw.piliszek@gmail.com <mailto:radoslaw.piliszek@gmail.com>> ha scritto: If one sets glance_file_datadir_volume to non-default, then glance-api gets deployed on all hosts.
-yoctozepto
On Fri, 19 Nov 2021 at 10:51, Ignazio Cassano <ignaziocassano@gmail.com <mailto:ignaziocassano@gmail.com>> wrote:
Hello Franck, glance is not deployed on all nodes at default. I got the same problem In my case I have 3 controllers. I created an nfs share on a storage server where to store images. Before deploying glance, I create /var/lib/glance/images on the 3 controllers and I mount the nfs share. This is my fstab on the 3 controllers:
10.102.189.182:/netappopenstacktst2_glance /var/lib/glance nfs rw,user=glance,soft,intr,noatime,nodiratime
In my globals.yml I have: glance_file_datadir_volume: "/var/lib/glance" glance_backend_file: "yes"
This means images are on /var/lib/glance and since it is a nfs share all my 3 controlles can share images. Then you must deploy. To be sure the glance container is started on all controllers, since I have 3 controllers, I deployed 3 times changing the order in the inventory. First time: [control] A B C
Second time: [control] B C A
Third time: [control] C B A
Or you can deploy glance 3 times using -t glance and -l <controllername>
As far as the instance stopped, I got I bug with a version of kolla. https://bugs.launchpad.net/kolla-ansible/+bug/1941706 <https://bugs.launchpad.net/kolla-ansible/+bug/1941706> Now is corrected and with kolla 12.2.0 it works. Ignazio
Il giorno mer 17 nov 2021 alle ore 23:17 Franck VEDEL <franck.vedel@univ-grenoble-alpes.fr <mailto:franck.vedel@univ-grenoble-alpes.fr>> ha scritto: