[Maas-devel] Bootstrap Errors in MAAS (1.6b6)

[Julian Edwards]

David,

IPs can get recycled quite quickly once a machine is released in MAAS (by 
default they are only allocated when a machine is in use).  If you're seeing 
this behaviour then nodes must be getting allocated and released very quickly.  
In my own juju testing things were fine so I've no idea why that would be.

There is an API call to force a pre-allocation though, which keeps the IP 
constant on a Node.  Try it out:

http://maas.ubuntu.com/docs/api.html#node
(search for claim_sticky_ip_address)

Let me know how you get on.

Julian.

On Monday 04 Aug 2014 20:58:57 David Britton wrote:
> I've had three failures now, all three have gotten the same IP addresses
> (but a different physical MAAS host each time):
> 
>     Waiting for address
>     Attempting to connect to <hostname here>:22
>     Attempting to connect to 10.1.3.229:22
>     Attempting to connect to 10.1.3.228:22
>     [...]
> 
>     2014-08-05 02:11:32 INFO juju.mongo open.go:95 dialled mongo
> successfully 2014-08-05 02:11:32 INFO juju.replicaset replicaset.go:67
> Initiating replicaset with config replicaset.Config{Name:"juju", Version:1,
> Members:[]replicaset.Member{replicaset.Member{Id:1,
> Address:"10.1.3.228:37017", Arbiter:(*bool)(nil),
> BuildIndexes:(*bool)(nil), Hidden:(*bool)(nil), Priority:(*float64)(nil),
> Tags:map[string]string{"juju-machine-id":"0"},
> SlaveDelay:(*time.Duration)(nil), Votes:(*int)(nil)}}} 2014-08-05 02:11:35
> WARNING juju.replicaset replicaset.go:86 Initiate: fetching replication
> status failed: cannot get replica set status: can't get
> local.system.replset config from self or any seed (EMPTYCONFIG) [... lots
> of this same error/warning pattern ...]
> 
>     2014-08-05 00:59:49 WARNING juju.replicaset replicaset.go:86 Initiate:
> fetching replication status failed: cannot get replica set status: can't
> get local.system.replset config from self or any seed (EMPTYCONFIG)
> 2014-08-05 00:59:50 INFO juju.worker.peergrouper initiate.go:94 finished
> MaybeInitiateMongoServer 2014-08-05 00:59:50 ERROR juju.cmd
> supercommand.go:323 cannot initiate replica set: cannot get replica set
> status: can't get local.system.replset config from self or any seed
> (EMPTYCONFIG) ERROR bootstrap failed: subprocess encountered error code 1
>     Stopping instance...
> 
> Notice the following entries are repeated *many* times in the leases
> file:
> 
>     host 10.1.3.229 {
>         dynamic;
>         hardware ethernet ec:a8:6b:fe:13:1b;
>         fixed-address 10.1.3.229;
>     }
>     host 10.1.3.229 {
>         dynamic;
>         deleted;
>     }
> 
> The probme: I seem to be hitting bootstrap problems in this MAAS.  I asked
> the juju guys a similar question, but not MAAS focused (just looking at
> that mongo replication stuff).  The frequency with which I'm repeating it
> now makes me think there is something in MAAS that is causing Juju to get
> upset - and Juju is handling it poorly.
> 
> 
> I'm not sure if this is related, but notice the very odd ip address
> allocation pattern for the 10.1.3.x range here -- this is really weird to
> me.  We have allocated the whole /24 for the 'static' range in maas.  The
> ":<number>" is how many times that address appears as a 'host <ip> {'
> pattern in the leases file.
> 
>     10.1.3.1:0
>     # [.. all zeros ..]
>     10.1.3.73:0
>     10.1.3.74:16
>     10.1.3.75:16
>     10.1.3.76:0
>     # [.. all zeros ..]
>     10.1.3.218:0
>     10.1.3.219:0
>     10.1.3.220:42
>     10.1.3.221:42
>     10.1.3.222:42
>     10.1.3.223:42
>     10.1.3.224:26
>     10.1.3.225:28
>     10.1.3.226:18
>     10.1.3.227:32
>     10.1.3.228:46
>     10.1.3.229:42
>     10.1.3.230:0
>     # [.. all zeros ..]
> 
> I took this one step further and looked at what MAC address was associated
> with which each host stanza.  Seems to be many for each.  In other words,
> it's not that the same host is associated with the same ip each time.
> 
>     http://paste.ubuntu.com/7957417/
> 
> Any insigts?