storm-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From saiprasad mishra <saiprasadmis...@gmail.com>
Subject Re: [SURVEY] What version of Storm are you using?
Date Sat, 20 Aug 2016 08:28:59 GMT
@Walmart E-Commerce

Upgraded one of our cluster from 0.10 to 1.0.1 a month back and running
successfully without any issues. Its worth all the time spent on the
upgrade as the new features are awesome for debugging and for everybody.

Another critical cluster is already upgraded from 0.9.4 to 1.0.1 and will
be live soon in matter of days.

Upgrade was very straight forward as most of the changes are in pom.xml and
we mainly use the storm-kafka spout with kafka 0.9.X client libs and kafka
0.9.X brokers.


1) Only blocker was log4j2 change as mentioned above. We were using
scribe-logback for logging.
So we ended up creating our own scribe log4j2 custom appender which was
anyway needed for any type of applications within our team.


2) One issue which we are facing right now is killing the topology from cli
some times does not kill the worker in 1.0.1 though the UI says
KILLED.Though there is a workaround for this(like clearing nimbus state by
hand) but still looking for a clean solution(considering to go to 1.0.2 for
this plus other stability fixes it has). I saw couple of JIRAs but not sure
whether those made it 1.0.2

Overall +1 for storm 1.0.x upgrade. Lot of useful features which makes
everybody's life easier especially ops team.

Regards
Sai




On Thu, Aug 18, 2016 at 9:17 PM, Zhechao Ma <mazhechaomaillist@gmail.com>
wrote:

> We upgrade from 0.9.5 to 1.0.1 recently.
>
> 2016-08-18 11:26 GMT+08:00 P. Taylor Goetz <ptgoetz@gmail.com>:
>
>> +1
>>
>> Any and all feedback is always welcome.
>>
>> -Taylor
>>
>> On Aug 17, 2016, at 9:22 PM, Jungtaek Lim <kabhwan@gmail.com> wrote:
>>
>> off-topic:
>> Andrew and Erik, please feel free to post the issue to dev@ whenever you
>> meet the issue from streamparse, storm-mesos, and etc. which might be cause
>> of upstream so that Storm dev. community can help if possible.
>>
>> 2016년 8월 18일 (목) 오전 10:07, Andrew Montalenti <andrew@parsely.com>님이
작성:
>>
>>> Erik, re: this issue, "blocked by us dealing with mysterious Storm
>>> Worker process heartbeat failures", do you have any topo's @ Groupon using
>>> ShellSpout (multi-lang)?
>>>
>>> If so, the patch/resolution in STORM-1928 (released in 1.0.2, see JIRA
>>> for details) may be helpful.
>>>
>>> On Aug 17, 2016 6:07 PM, "Jungtaek Lim" <kabhwan@gmail.com> wrote:
>>>
>>>> Side note for users waiting stable version of 1.x: 1.0.2 is worth to
>>>> try out since it's well tested, and fixes various critical bugs users met.
>>>>
>>>> - Jungtaek Lim (HeartSaVioR)
>>>>
>>>> 2016년 8월 18일 (목) 오전 6:58, Erik Weathers <eweathers@groupon.com>님이
작성:
>>>>
>>>>> Groupon is using 0.9.6.
>>>>>
>>>>> Switching to 0.10.0 or 1.0+ will take us more time largely because of
>>>>> the "logback to log4j 2" change in 0.10.  We have a bunch of internal
teams
>>>>> using a library for logging which is based on logback, and it's definitely
>>>>> going to be herding cats to get them to upgrade their topologies.
>>>>>
>>>>> We are also responsible for the storm-mesos framework, which we will
>>>>> hopefully come back to adding support for 1.0+ next month on.   Spending
>>>>> time on that is currently blocked by us dealing with mysterious Storm
>>>>> Worker process heartbeat failures.
>>>>>
>>>>> - Erik
>>>>>
>>>>> On Wed, Aug 17, 2016 at 2:47 PM, Joaquin Menchaca <
>>>>> jmenchaca@gobalto.com> wrote:
>>>>>
>>>>>> 0.10.0
>>>>>>
>>>>>
>>>>>
>

Mime
View raw message