cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rohit Yadav <>
Subject Re: DevCloud2 secondary storage vm
Date Thu, 10 Jan 2013 01:46:35 GMT

On 09-Jan-2013, at 5:10 PM, Marcus Sorensen <> wrote:

> Oh, I see. No, it's defaulting to host = Yes, if you're running
> the management server in the devcloud it probably resolves the devcloud
> host to the xenbr1 IP? In which case it doesn't make sense to edit the
> devcloud.cfg.

Yes, I get your point, yes the problem is with a bug in how the mgmt starts. The first time
it starts it reads the file to get the cluster node ip which is misleading name,
but it's the mgmt server host IP which is set to localhost if no IP on nic is found. It would
use the first gateway's IP which is nat in case of DevCloud so you get

To fix this issue, just make a (.override, forgot if this still works) file
that has cluster node ip equal to and if I recall correctly despite the database
host value or devcloud.cfg host value, it would stick with host ip =

Lastly, even though we add the host field for global param in devcloud.cfg there is an issue
(why I removed it and put it in devcloud.sql), it would require you to restart the mgmt server
before system vms are deployed, else the host ip (ie. mgmt server ip) will be sent as
to the systemvms and the agent may fail to connect back to the mgmt server and the state will
hang as "Starting".

> I like to have everything in the devcloud in order to keep it a tidy,
> portable, replicatable package. Then I connect my IDE to the sourcecode via
> NFS or SSHFS in the devcloud.

We should create a DevCloud appliance with a desktop env and preconfigured IDE and tools.

One method I'd explored: you can have a rsync daemon to sync files from your local repo on
host os to location in devcloud, (NFS or any other fs will be slow) run mgmt server in devcloud
and use IDE on host os to connect to debug socket on on port 8787.


> On Wed, Jan 9, 2013 at 5:57 PM, Rohit Yadav <> wrote:
>> What we should do is have an overrides option in maven itself to override
>> the default cfg file with a developer's own config file.
>> The reason it is set to is that most developers (or users if
>> they are trying to checkout cloudstack) workflow would be to run mgmt
>> server and db server on their host os and use DevCloud just as recycle-able
>> host+storage.
>> Not objecting, just that we should fix it in maven so if a
>> devcloud.cfg.override is available (or some other name), it would read that
>> file only while deploysrvr is done.
>> Regards.
>> On 09-Jan-2013, at 4:11 PM, Marcus Sorensen <> wrote:
>>> Unless anyone objects, I'm going to change the devcloud.cfg for marvin to
>>> add as the 'host' global config. Without this, you don't
>> get
>>> the tinylinux template. With it, and a restart after applying the config,
>>> you get the template as soon as the secondary storage vm is provisioned.
>>> On Sun, Jan 6, 2013 at 8:55 PM, Marcus Sorensen <>
>> wrote:
>>>> I had the exact same thing as well following devcloud2 instructions. It
>>>> seems the ssvm only connects to the host-only network. That's probably
>> fine
>>>> for basic testing since there is already a tinylinux template, but we
>>>> should consider setting the default zone up such that the ssvm will
>> connect
>>>> to the NAT so it can get DNS/internet for actual functional testing
>>>> On Sun, Jan 6, 2013 at 2:36 PM, Sebastien Goasguen <
>>> wrote:
>>>>> Jamshid, I don't think anyone answered you ?
>>>>> Are you having issues starting instances ?
>>>>> -Sebastien
>>>>> On Dec 28, 2012, at 8:50 PM, Jamshid Afshar <
>>>>> wrote:
>>>>>> I built and am running the management server inside the devcloud2
>>>>>> I had to change "host" in Global Settings to, do I
>>>>>> to change other settings?
>>>>>> Maybe this expected, but I see this error on Secondary Storage system
>>>>>> vm "s-1-vm" in /var/log/cloud/systemvm.log:
>>>>>> WARN NfsSecondaryStorageResource Error in configuring route to
>>>>>> internal ip err=RTNELINK answers: Invalid Argument
>>>>>> The script /usr/local/cloud/systemvm/ reports "ERROR:
>>>>>> not resolving". If I comment the "exit" out the
>>>>>> other tests succeed (ie, it sees management server at
>>>>>> Should a DNS server be running on my (all-in-one) management server
>>>>>> host
>>>>>> Should I be able to ping from s-1-vm? I can't. "route"
>>>>> shows:
>>>>>> * eth1
>>>>>> * eth2
>>>>>> * eth3
>>>>>> link-local *     eth0
>>>>>> default   eth2
>>>>>> Thanks,
>>>>>> Jam

View raw message