cloudstack-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Andrija Panic <andrija.pa...@gmail.com>
Subject Re: SolidFire with CloudStack 4.11.3 and vsphere 6.5
Date Fri, 13 Sep 2019 21:54:42 GMT
Christian,

In the meantime, you can try the *SolidFireShared *plugin, which I just
tested and it seems to works fine - this means one huge LUN/datastore for
hosting all ACS volume (you can make Primary Storage per cluster)
(FYI, this is different than the 1-2-1 "SolidFire" plugin in the sense that
that one makes 1 LUN, mount it as Datastore and creates only 1 ACS volume
on it. This is repeated for EVERY ACS volume you created - this means you
have the limit of whatever is the max number of datastores in VMware!)

You still do in vCenter
1) create iSCSI adapter in vCenter
2) do network binding

create StoragePool scope=cluster
zoneid=af61811f-3ca6-4927-ab0d-5bb6d693e3e7
podid=954065ed-a173-4c52-9f6f-062cd9b17ddb
clusterid=72750371-a6ce-4d97-b567-1a9aefc416f8 name=SFSHARED
provider=SolidFireShared managed=false capacityBytes=107374182400
capacityIops=15000
url="MVIP=10.0.1.40;SVIP=10.254.254.10;clusterAdminUsername=admin;clusterAdminPassword=password;minIops=15000;maxIops=70000;burstIops=75000;datacenter=Trillian"
tags=SFSHARED

^^^ CMK command, you can work out the GUI values from that.

Note that the min_iops can NOT be 50.000 as in that example (it will fail),
you can use 15.000 at most (again, this is only for the min_iops values
(and the"capacityIops" which needs to have same values as the min_iops in
the URL).
max_iops can be at most 200.000 (both these numbers are the limitations of
the SF/ElementOS, not from ACS)

Let me know if more help needed.

Cheers
Andrija

On Fri, Sep 13, 2019, 08:38 Andrija Panic <andrija.panic@gmail.com> wrote:

> Cool - would be happy to learn that the issue is solved - want to make
> after that the proper documentation upstream.
>
> Thx Mike!
>
> On Fri, Sep 13, 2019, 07:19 Tutkowski, Mike <Mike.Tutkowski@netapp.com>
> wrote:
>
>> Hi Christian,
>>
>> I work at NetApp. I can contact you off list to give you instructions on
>> how to get support at NetApp involved.
>>
>> Thanks, Andrija, for debugging!
>>
>> Talk to you later,
>> Mike
>>
>> On 9/13/19, 1:15 AM, "christian.kirmse@zv.fraunhofer.de" <
>> christian.kirmse@zv.fraunhofer.de> wrote:
>>
>>     NetApp Security WARNING: This is an external email. Do not click
>> links or open attachments unless you recognize the sender and know the
>> content is safe.
>>
>>
>>
>>
>>     Hi Andrija,
>>
>>     thanks for your effort, that is some good and bad news at once.
>>     The good thing is, we are not too stupid to set it up.
>>
>>     I will get in contact with NetApp and hear what they say, I’m afraid
>> it gonna take some time…
>>
>>     Regards
>>     Christian
>>
>>     > On 13. Sep 2019, at 00:00, Andrija Panic <andrija.panic@gmail.com>
>> wrote:
>>     >
>>     > Right,
>>     >
>>     > so I managed to reproduce your issues, 2 times, using ACS
>> 4.13/master and
>>     > VMware 6.5... got the same error message, after a lot of tasks being
>>     > executed on VMware side.
>>     >
>>     > The steps I take were to (first try) don't even add the iSCSI
>> Sotfware
>>     > adapter, but then when trying to spin a VM, there is obviously no
>> IQN
>>     > identifier assigned to the ESXi hosts, and thus not part of the DB
>> (hosts
>>     > table) - so, as expected, you have added the SF to ACS (SolidFire
>> plugin,
>>     > Managed, proper URL defined), but it fails to spin a VM.
>>     >
>>     > Clean up everything in ACS, wipe the Primary Storage. Add iSCSI
>> Software
>>     > adapter in vCenter/ESXi hosts, configure proper binding to
>> vSwitchXXX and
>>     > then add again SF (again, SolidFire provider, Managed and proper
>> ULR) to
>>     > SF, try to spin a VM.
>>     > Now, a series of things are happening in vCenter.
>>     >
>>     > -  Static iSCSI target is added to ESXI hosts,
>>     > - Rescanning HBAs,
>>     > -  Creating datastore same size as the volume/template itself
>>     > - Deploying OVF template
>>     > - Unregistering VM
>>     > - Moving files around
>>     > - unmounting VMFS
>>     > - Removing iSCSI static targets
>>     > - Rescan HBA
>>     > - Again adding iSCSI static targets
>>     > - Rescan HBAs
>>     > - Rescan VMFS
>>     > - RENAME datastore
>>     > - unmount datastore
>>     > - remove iSCSI targets.
>>     >
>>     > The error from the ACS is:
>>     > message: Datastore '-iqn.2010-01.com.solidfire:hl1k.root-32.29-0'
>> is not
>>     > accessible. No connected and accessible host is attached to this
>> datastore
>>     >
>>     > The problem is - this datastore (in it's latest, renamed state) - is
>>     > unmounted from ESXI hosts, but can't be removed, NOR can I mount it
>> - I get
>>     > the vCenter message of "Operation failed, diagnostics report:
>> Unable to
>>     > find volume uuid[5d7abd9a-273aa9d5-bffe-1e00d4010711] lvm
>>     > [snap-329aa3ea-5d7abd01-a5c83210-c87c-1e00d4010711] devices"
>>     >
>>     > So something is broken here...
>>     >
>>     > Will try other scenarios soon (SolidFire Shared, etc).
>>     >
>>     >
>>     >
>>     >
>>     >
>>     >
>>     >
>>     > On Thu, 5 Sep 2019 at 11:33, Andrija Panic <andrija.panic@gmail.com>
>> wrote:
>>     >
>>     >> That sounds OK to me, the steps to add SF. That should create a
>> datastore
>>     >> per each created volume you create (if not mistaken). The other
>> way is to
>>     >> use SolidFireShared plugin, which should create a single datastore
>> and
>>     >> place all volumes in it (datastore=LUN=single SF volume).
>>     >>
>>     >> Can you please answer my question from previous email, and also
>> you can
>>     >> chech for datastore statuses in vCenter for any error, something
>> is not
>>     >> right...
>>     >>
>>     >> Andrija
>>     >>
>>     >> On Thu, Sep 5, 2019, 16:15 <christian.kirmse@zv.fraunhofer.de>
>> wrote:
>>     >>
>>     >>> Hi Andrija,
>>     >>>
>>     >>>> On 5. Sep 2019, at 15:07, Andrija Panic <andrija.panic@gmail.com
>> >
>>     >>> wrote:
>>     >>>
>>     >>>> the message is that no host is connected to that specific
>> datastore -
>>     >>>> "Unable to start VM on Host[-1-Routing] due to StartCommand
>> failed due
>>     >>> to
>>     >>>> Exception: java.lang.RuntimeException
>>     >>>> Message: Datastore
>> '-iqn.2010-01.com.solidfire:x64j.root-29.17-0' is not
>>     >>>> accessible. ***No connected and accessible host is attached
to
>> this
>>     >>>> datastore***."
>>     >>>>
>>     >>>> You can see that message being returned by VMware actually,
not
>> ACS (I
>>     >>>> checked the code for that message - no results)
>>     >>>>
>>     >>>
>> https://vmninja.wordpress.com/2019/04/05/remove-inaccessible-datastore-from-inventory/
>>     >>>>
>>     >>> I assumed the message did come from ACS since there were no
>> corresponding
>>     >>> messages in vcenter.
>>     >>> Before every new deployment attempt I get sure there are not
>> leftovers,
>>     >>> neither in vcenter nor on the solidfire.
>>     >>>
>>     >>>>
>>     >>>> Can you describe exactly how did you add SF to the ACS/VMware
-
>> you
>>     >>> already
>>     >>>> wrote you created iSCSI HBAs....? What are the
>> parameters/options used
>>     >>> to
>>     >>>> add SF as Primary Storage to ACS? I expect (since no proper
>>     >>> documentation
>>     >>>> yet) that you might have somehow incomplete or wrong setup in
>> place.
>>     >>> In vcenter there was already a VMkernel interface which is used
>> for nfs
>>     >>> datastrores.
>>     >>> So first I created a ISCSI Software Adapter and added the existing
>>     >>> VMkernel interface via the network port binding option to the
>> ISCSI
>>     >>> Software Adapter.
>>     >>> Afterwards I did a force re-connect of the host in ACS.
>>     >>>
>>     >>> Then I followed the youtube guide by Mike to add the solidfire as
>> primary
>>     >>> storage with
>>     >>> Protocol = custom
>>     >>> Provider = solidfire
>>     >>> Managed = true
>>     >>> Filled in IOPS and Capacity and the URL as follows:
>>     >>>
>>     >>>
>> MVIP=<MVIP>;SVIP=<SVIP>;clusterAdminUsername=<USERNAME>;clusterAdminPassword=<PASSWORD>;clusterDefaultMinIops=100;clusterDefaultMaxIops=200;clusterDefaultBurstIopsPercentOfMaxIops=1.5
>>     >>>
>>     >>> The odd thing is that I managed to have one deployment working,
>> but only
>>     >>> once.
>>     >>> This VM was running fine, I could ssh to it, use the console
>> etc...
>>     >>>
>>     >>> Regards
>>     >>> Christian
>>     >>>> On Thu, 5 Sep 2019 at 13:29, <christian.kirmse@zv.fraunhofer.de>
>> wrote:
>>     >>>>
>>     >>>>> Thanks for taking time to look into it.
>>     >>>>>
>>     >>>>> https://pastebin.com/utPhEVkW
>>     >>>>>
>>     >>>>> Regards
>>     >>>>> Christian
>>     >>>>>> On 5. Sep 2019, at 13:16, Andrija Panic <
>> andrija.panic@gmail.com>
>>     >>> wrote:
>>     >>>>>>
>>     >>>>>> Can you share the mgmt logs when the problem happens?
Please
>> upload to
>>     >>>>>> pastebin or similar.
>>     >>>>>>
>>     >>>>>> Andrija
>>     >>>>>>
>>     >>>>>> On Thu, 5 Sep 2019 at 11:45, <
>> christian.kirmse@zv.fraunhofer.de>
>>     >>> wrote:
>>     >>>>>>
>>     >>>>>>> Hi,
>>     >>>>>>>
>>     >>>>>>> I have managed to overcome the problem by forcing
Cloudstack
>> to
>>     >>>>> reconnect
>>     >>>>>>> the host after I configured the iSCSI HBA.
>>     >>>>>>> It seems that Cloudstack also scans for such capabilities
>> during the
>>     >>>>>>> reconnect.
>>     >>>>>>>
>>     >>>>>>> But now I have trouble deploying VMs on the storage,
>> sometimes it is
>>     >>>>>>> successful and sometimes not, I was not able to
find a
>> pattern.
>>     >>>>>>> If the deployment fails Cloudstack says that the
IQN was not
>>     >>> reachable
>>     >>>>> by
>>     >>>>>>> the host, the odd part is that there is no such
message in
>> vcenter
>>     >>> so it
>>     >>>>>>> seems there was no attempt to attach the storage
to the ESXi…
>>     >>>>>>>
>>     >>>>>>> Does anyone know this kind of issue?
>>     >>>>>>>
>>     >>>>>>> Regards
>>     >>>>>>> Christian
>>     >>>>>>>
>>     >>>>>>>
>>     >>>>>>> On 2019/09/04 14:13:36, <c...@zv.fraunhofer.de>
wrote:
>>     >>>>>>>> Hi,>
>>     >>>>>>>>
>>     >>>>>>>>
>>     >>>>>>>>
>>     >>>>>>>> we are currently doing a PoC with SolidFire
and CloudStack
>> and
>>     >>> trying
>>     >>>>> to
>>     >>>>>>> figure out if it’s a fitting solution for our
use cases.>
>>     >>>>>>>>
>>     >>>>>>>>
>>     >>>>>>>>
>>     >>>>>>>> But I am stuck at the point when CloudStack
tries to create
>> a VM on
>>     >>> the
>>     >>>>>>> solid fire storage.>
>>     >>>>>>>>
>>     >>>>>>>> I can see that it has already copied the template
to a
>> SolidFire
>>     >>> Volume
>>     >>>>>>> but then the error message "Not all hosts in the
compute
>> cluster
>>     >>> support
>>     >>>>>>> iSCSI.” appears in the logs.>
>>     >>>>>>>>
>>     >>>>>>>>
>>     >>>>>>>>
>>     >>>>>>>> On the ESXi I have created a iSCSI HBA and attached
it to a
>> VMKernel
>>     >>>>>>> adapter, is there anything else to do?>
>>     >>>>>>>>
>>     >>>>>>>> Is there any documentation for the setup?>
>>     >>>>>>>>
>>     >>>>>>>> I have only found the youtube videos by Mike,
but they does
>> not
>>     >>> focus
>>     >>>>> on
>>     >>>>>>> the vsphere setup part.>
>>     >>>>>>>>
>>     >>>>>>>>
>>     >>>>>>>>
>>     >>>>>>>> Regards>
>>     >>>>>>>>
>>     >>>>>>>> Christian>
>>     >>>>>>>
>>     >>>>>>>
>>     >>>>>>> --
>>     >>>>>>> Christian Kirmse
>>     >>>>>>> Fraunhofer-Gesellschaft e.V. / Zentrale
>>     >>>>>>> Abteilung C7 Kommunikationsmanagement
>>     >>>>>>> Schloss Birlinghoven, 53754 Sankt Augustin
>>     >>>>>>> Tel: (+49 2241) 14-2719
>>     >>>>>>> Fax: (+49 2241) 144-2719
>>     >>>>>>> mailto:christian.kirmse@zv.fraunhofer.de
>>     >>>>>>> http://www.fraunhofer.de
>>     >>>>>>>
>>     >>>>>>>
>>     >>>>>>
>>     >>>>>> --
>>     >>>>>>
>>     >>>>>> Andrija Panić
>>     >>>>>
>>     >>>>>
>>     >>>>
>>     >>>> --
>>     >>>>
>>     >>>> Andrija Panić
>>     >>>
>>     >>>
>>     >
>>     > --
>>     >
>>     > Andrija Panić
>>
>>
>>
>>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message