tasks reassigning when a worker node downs and recover

[Miyakawa Takeshi <fj...@gmail.com>]

Hi, i am new to storm and i met a problem with tasks reassigning when
a worker node downs and recover.

worker node 1-5 running my cluster.
i submit a topology to my cluster, and nimbus assign some task to node 1.
when node 1 is down (whole machine level), supervisor and all worker
processes down too,
tasks assigned to node 1 are reassign to other nodes. it's OK.

but, when node 1 reboot and supervisor has start, tasks (reassign to
other nodes) return to node 1!

i don't want to such returning.

so, i have two questions,

Q1. such returning is proper behavior to Apache Storm?

Q2. if answer of Q1 is yes, does exist some ways to prevent such returning?

i am a Japanese and poor at English, sorry.

i use Apache Storm 0.9.1 now.

Re: tasks reassigning when a worker node downs and recover

["UmaraDissa1 ." <um...@gmail.com>]

I didn't notice anything in the release notes of 0.9.2 (
https://storm.apache.org/2014/06/25/storm092-released.html) which would
indicate any change of this sort. Also, the version I am using is
0.9.2-incubating as well.

Do keep us updated though

On Fri, Jan 9, 2015 at 3:53 PM, Miyakawa Takeshi <fj...@gmail.com>
wrote:

> Thanks you.
>
> I'll try to pursuit the behavior of Apache Storm.
> I am going to migrate to 0.9.2 from 0.9.1.
> So, the behavior may be changed.
> I'll report this problem in some weeks.
>
> 2015-01-08 17:23 GMT+09:00 UmaraDissa1 . <um...@gmail.com>:
> > Hi Miyakawa,
> >
> > I don't think any tasks that were reassigned to a different node due a
> node
> > failure would be dynamically assigned back to the original node that
> failed
> > when that node comes back up again.
> >
> > I haven't observed this, but take this with a grain of salt as I am a
> bit of
> > a newbie at Storm myself.
> >
> >
> > Cheers,
> > Kosala.
> >
> > On Mon, Jan 5, 2015 at 9:04 PM, Miyakawa Takeshi <fj...@gmail.com>
> > wrote:
> >>
> >> Hi, i am new to storm and i met a problem with tasks reassigning when
> >> a worker node downs and recover.
> >>
> >> worker node 1-5 running my cluster.
> >> i submit a topology to my cluster, and nimbus assign some task to node
> 1.
> >> when node 1 is down (whole machine level), supervisor and all worker
> >> processes down too,
> >> tasks assigned to node 1 are reassign to other nodes. it's OK.
> >>
> >> but, when node 1 reboot and supervisor has start, tasks (reassign to
> >> other nodes) return to node 1!
> >>
> >> i don't want to such returning.
> >>
> >> so, i have two questions,
> >>
> >> Q1. such returning is proper behavior to Apache Storm?
> >>
> >> Q2. if answer of Q1 is yes, does exist some ways to prevent such
> >> returning?
> >>
> >> i am a Japanese and poor at English, sorry.
> >>
> >> i use Apache Storm 0.9.1 now.
> >
> >
>

Re: tasks reassigning when a worker node downs and recover

[Miyakawa Takeshi <fj...@gmail.com>]

Thanks you.

I'll try to pursuit the behavior of Apache Storm.
I am going to migrate to 0.9.2 from 0.9.1.
So, the behavior may be changed.
I'll report this problem in some weeks.

2015-01-08 17:23 GMT+09:00 UmaraDissa1 . <um...@gmail.com>:
> Hi Miyakawa,
>
> I don't think any tasks that were reassigned to a different node due a node
> failure would be dynamically assigned back to the original node that failed
> when that node comes back up again.
>
> I haven't observed this, but take this with a grain of salt as I am a bit of
> a newbie at Storm myself.
>
>
> Cheers,
> Kosala.
>
> On Mon, Jan 5, 2015 at 9:04 PM, Miyakawa Takeshi <fj...@gmail.com>
> wrote:
>>
>> Hi, i am new to storm and i met a problem with tasks reassigning when
>> a worker node downs and recover.
>>
>> worker node 1-5 running my cluster.
>> i submit a topology to my cluster, and nimbus assign some task to node 1.
>> when node 1 is down (whole machine level), supervisor and all worker
>> processes down too,
>> tasks assigned to node 1 are reassign to other nodes. it's OK.
>>
>> but, when node 1 reboot and supervisor has start, tasks (reassign to
>> other nodes) return to node 1!
>>
>> i don't want to such returning.
>>
>> so, i have two questions,
>>
>> Q1. such returning is proper behavior to Apache Storm?
>>
>> Q2. if answer of Q1 is yes, does exist some ways to prevent such
>> returning?
>>
>> i am a Japanese and poor at English, sorry.
>>
>> i use Apache Storm 0.9.1 now.
>
>

Re: tasks reassigning when a worker node downs and recover

["UmaraDissa1 ." <um...@gmail.com>]

Hi Miyakawa,

I don't think any tasks that were reassigned to a different node due a node
failure would be dynamically assigned back to the original node that failed
when that node comes back up again.

I haven't observed this, but take this with a grain of salt as I am a bit
of a newbie at Storm myself.


Cheers,
Kosala.

On Mon, Jan 5, 2015 at 9:04 PM, Miyakawa Takeshi <fj...@gmail.com>
wrote:

> Hi, i am new to storm and i met a problem with tasks reassigning when
> a worker node downs and recover.
>
> worker node 1-5 running my cluster.
> i submit a topology to my cluster, and nimbus assign some task to node 1.
> when node 1 is down (whole machine level), supervisor and all worker
> processes down too,
> tasks assigned to node 1 are reassign to other nodes. it's OK.
>
> but, when node 1 reboot and supervisor has start, tasks (reassign to
> other nodes) return to node 1!
>
> i don't want to such returning.
>
> so, i have two questions,
>
> Q1. such returning is proper behavior to Apache Storm?
>
> Q2. if answer of Q1 is yes, does exist some ways to prevent such returning?
>
> i am a Japanese and poor at English, sorry.
>
> i use Apache Storm 0.9.1 now.
>