You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@kafka.apache.org by Dmitriy Vsekhvalnov <dv...@gmail.com> on 2017/07/07 11:47:37 UTC
Where to run kafka-consumer-groups.sh from?
Hi all,
question about lag checking. We've tried to periodically sample consumer
lag with:
kafka-consumer-groups.sh --bootstrap-server broker:9092 --new-consumer
--group service-group --describe
it's all fine, but depending on host we run it from it gives different
results.
E.g:
- when running from one of the broker hosts itself we getting close to 0
figures.
- when running from remote host, we getting 30-60 in average (i suspect
there are multiple remote calls to broker involved, so difference due to
timing).
My question is what is correct way to use it? From broker host itself?
Re: Where to run kafka-consumer-groups.sh from?
Posted by Dmitriy Vsekhvalnov <dv...@gmail.com>.
Nothing painful really at this point.
We were just evaluating kafka, and noticed the discrepancy. Was wondering
what's the 'design way' of using that tool.
Unlikely we gonna use it for prod monitoring at all. Just for initial
experiments.
On Tue, Jul 11, 2017 at 1:23 PM, Jan Filipiak <Ja...@trivago.com>
wrote:
> Hi,
>
> very likely due to timing. What problem is it causing you exactly that you
> want to work around?
> These differences shouldn't concern you to much I guess.
>
> We use the tool across continents and don't worry about it to much. Offset
> Commit interval makes everything blury anyways. If you can specify your
> pain more precisely maybe we can work around it.
>
> Best Jan
>
>
> On 10.07.2017 10:31, Dmitriy Vsekhvalnov wrote:
>
>> Guys, let me up this one again. Still looking for comments about
>> kafka-consumer-groups.sh
>> tool.
>>
>> Thank you.
>>
>> On Fri, Jul 7, 2017 at 3:14 PM, Dmitriy Vsekhvalnov <
>> dvsekhvalnov@gmail.com>
>> wrote:
>>
>> I've tried 3 brokers on command line, like that:
>>>
>>> /usr/local/kafka/bin/kafka-consumer-groups.sh --bootstrap-server
>>> broker:9092,broker_2:9092,broker_3:9092 --new-consumer --group
>>> logging-svc --describe
>>>
>>> it doesn't make any difference, still x10 times difference in figures
>>> when
>>> running on broker host vs. remote
>>>
>>> Here is snippet from console output (are you looking something specific
>>> in
>>> it? it looks normal a far as i can say):
>>>
>>>
>>> TOPIC PARTITION CURRENT-OFFSET LOG-END-OFFSET LAG
>>> CONSUMER-ID
>>>
>>> test.topic 54 4304 4309 35
>>> consumer-26-21f5050c-a43c-4254-bfcf-42e17dbdb651
>>>
>>> test.topic 40 4426 4436 10
>>> consumer-21-24f3ebca-004f-4aac-a348-638c9c6a02f0
>>>
>>> test.topic 59 4414 4420 63
>>> consumer-27-ed34f1b3-1be9-422b-bb07-e3c9913195c7
>>>
>>> test.topic 42 4389 4403 76
>>> consumer-22-75c2fc0a-5d5c-472d-b27e-e873030f82b6
>>>
>>> test.topic 27 4416 4422 24
>>> consumer-18-3be20568-8dd3-4679-a008-0ca64d31083c
>>>
>>>
>>>
>>>
>>> On Fri, Jul 7, 2017 at 2:52 PM, M. Manna <ma...@gmail.com> wrote:
>>>
>>> kafka-consumer-groups.sh --bootstrap-server broker:9092 --new-consumer
>>>> --group service-group --describe
>>>>
>>>> how many brokers do you have in the cluster? if you have more than one,
>>>> list them all using a comma csv with --bootstrap-server.
>>>>
>>>> Also, could you paste some results from the console printout?
>>>>
>>>> On 7 July 2017 at 12:47, Dmitriy Vsekhvalnov <dv...@gmail.com>
>>>> wrote:
>>>>
>>>> Hi all,
>>>>>
>>>>> question about lag checking. We've tried to periodically sample
>>>>> consumer
>>>>> lag with:
>>>>>
>>>>> kafka-consumer-groups.sh --bootstrap-server broker:9092 --new-consumer
>>>>> --group service-group --describe
>>>>>
>>>>> it's all fine, but depending on host we run it from it gives different
>>>>> results.
>>>>>
>>>>> E.g:
>>>>>
>>>>> - when running from one of the broker hosts itself we getting close
>>>>>
>>>> to 0
>>>>
>>>>> figures.
>>>>>
>>>>> - when running from remote host, we getting 30-60 in average (i
>>>>>
>>>> suspect
>>>>
>>>>> there are multiple remote calls to broker involved, so difference due
>>>>> to
>>>>> timing).
>>>>>
>>>>>
>>>>> My question is what is correct way to use it? From broker host itself?
>>>>>
>>>>>
>>>
>
Re: Where to run kafka-consumer-groups.sh from?
Posted by Jan Filipiak <Ja...@trivago.com>.
Hi,
very likely due to timing. What problem is it causing you exactly that
you want to work around?
These differences shouldn't concern you to much I guess.
We use the tool across continents and don't worry about it to much.
Offset Commit interval makes everything blury anyways. If you can
specify your pain more precisely maybe we can work around it.
Best Jan
On 10.07.2017 10:31, Dmitriy Vsekhvalnov wrote:
> Guys, let me up this one again. Still looking for comments about
> kafka-consumer-groups.sh
> tool.
>
> Thank you.
>
> On Fri, Jul 7, 2017 at 3:14 PM, Dmitriy Vsekhvalnov <dv...@gmail.com>
> wrote:
>
>> I've tried 3 brokers on command line, like that:
>>
>> /usr/local/kafka/bin/kafka-consumer-groups.sh --bootstrap-server
>> broker:9092,broker_2:9092,broker_3:9092 --new-consumer --group
>> logging-svc --describe
>>
>> it doesn't make any difference, still x10 times difference in figures when
>> running on broker host vs. remote
>>
>> Here is snippet from console output (are you looking something specific in
>> it? it looks normal a far as i can say):
>>
>>
>> TOPIC PARTITION CURRENT-OFFSET LOG-END-OFFSET LAG
>> CONSUMER-ID
>>
>> test.topic 54 4304 4309 35
>> consumer-26-21f5050c-a43c-4254-bfcf-42e17dbdb651
>>
>> test.topic 40 4426 4436 10
>> consumer-21-24f3ebca-004f-4aac-a348-638c9c6a02f0
>>
>> test.topic 59 4414 4420 63
>> consumer-27-ed34f1b3-1be9-422b-bb07-e3c9913195c7
>>
>> test.topic 42 4389 4403 76
>> consumer-22-75c2fc0a-5d5c-472d-b27e-e873030f82b6
>>
>> test.topic 27 4416 4422 24
>> consumer-18-3be20568-8dd3-4679-a008-0ca64d31083c
>>
>>
>>
>>
>> On Fri, Jul 7, 2017 at 2:52 PM, M. Manna <ma...@gmail.com> wrote:
>>
>>> kafka-consumer-groups.sh --bootstrap-server broker:9092 --new-consumer
>>> --group service-group --describe
>>>
>>> how many brokers do you have in the cluster? if you have more than one,
>>> list them all using a comma csv with --bootstrap-server.
>>>
>>> Also, could you paste some results from the console printout?
>>>
>>> On 7 July 2017 at 12:47, Dmitriy Vsekhvalnov <dv...@gmail.com>
>>> wrote:
>>>
>>>> Hi all,
>>>>
>>>> question about lag checking. We've tried to periodically sample consumer
>>>> lag with:
>>>>
>>>> kafka-consumer-groups.sh --bootstrap-server broker:9092 --new-consumer
>>>> --group service-group --describe
>>>>
>>>> it's all fine, but depending on host we run it from it gives different
>>>> results.
>>>>
>>>> E.g:
>>>>
>>>> - when running from one of the broker hosts itself we getting close
>>> to 0
>>>> figures.
>>>>
>>>> - when running from remote host, we getting 30-60 in average (i
>>> suspect
>>>> there are multiple remote calls to broker involved, so difference due to
>>>> timing).
>>>>
>>>>
>>>> My question is what is correct way to use it? From broker host itself?
>>>>
>>
Re: Where to run kafka-consumer-groups.sh from?
Posted by Dmitriy Vsekhvalnov <dv...@gmail.com>.
Guys, let me up this one again. Still looking for comments about
kafka-consumer-groups.sh
tool.
Thank you.
On Fri, Jul 7, 2017 at 3:14 PM, Dmitriy Vsekhvalnov <dv...@gmail.com>
wrote:
> I've tried 3 brokers on command line, like that:
>
> /usr/local/kafka/bin/kafka-consumer-groups.sh --bootstrap-server
> broker:9092,broker_2:9092,broker_3:9092 --new-consumer --group
> logging-svc --describe
>
> it doesn't make any difference, still x10 times difference in figures when
> running on broker host vs. remote
>
> Here is snippet from console output (are you looking something specific in
> it? it looks normal a far as i can say):
>
>
> TOPIC PARTITION CURRENT-OFFSET LOG-END-OFFSET LAG
> CONSUMER-ID
>
> test.topic 54 4304 4309 35
> consumer-26-21f5050c-a43c-4254-bfcf-42e17dbdb651
>
> test.topic 40 4426 4436 10
> consumer-21-24f3ebca-004f-4aac-a348-638c9c6a02f0
>
> test.topic 59 4414 4420 63
> consumer-27-ed34f1b3-1be9-422b-bb07-e3c9913195c7
>
> test.topic 42 4389 4403 76
> consumer-22-75c2fc0a-5d5c-472d-b27e-e873030f82b6
>
> test.topic 27 4416 4422 24
> consumer-18-3be20568-8dd3-4679-a008-0ca64d31083c
>
>
>
>
> On Fri, Jul 7, 2017 at 2:52 PM, M. Manna <ma...@gmail.com> wrote:
>
>> kafka-consumer-groups.sh --bootstrap-server broker:9092 --new-consumer
>> --group service-group --describe
>>
>> how many brokers do you have in the cluster? if you have more than one,
>> list them all using a comma csv with --bootstrap-server.
>>
>> Also, could you paste some results from the console printout?
>>
>> On 7 July 2017 at 12:47, Dmitriy Vsekhvalnov <dv...@gmail.com>
>> wrote:
>>
>> > Hi all,
>> >
>> > question about lag checking. We've tried to periodically sample consumer
>> > lag with:
>> >
>> > kafka-consumer-groups.sh --bootstrap-server broker:9092 --new-consumer
>> > --group service-group --describe
>> >
>> > it's all fine, but depending on host we run it from it gives different
>> > results.
>> >
>> > E.g:
>> >
>> > - when running from one of the broker hosts itself we getting close
>> to 0
>> > figures.
>> >
>> > - when running from remote host, we getting 30-60 in average (i
>> suspect
>> > there are multiple remote calls to broker involved, so difference due to
>> > timing).
>> >
>> >
>> > My question is what is correct way to use it? From broker host itself?
>> >
>>
>
>
Re: Where to run kafka-consumer-groups.sh from?
Posted by Dmitriy Vsekhvalnov <dv...@gmail.com>.
I've tried 3 brokers on command line, like that:
/usr/local/kafka/bin/kafka-consumer-groups.sh --bootstrap-server
broker:9092,broker_2:9092,broker_3:9092 --new-consumer --group logging-svc
--describe
it doesn't make any difference, still x10 times difference in figures when
running on broker host vs. remote
Here is snippet from console output (are you looking something specific in
it? it looks normal a far as i can say):
TOPIC PARTITION CURRENT-OFFSET LOG-END-OFFSET LAG
CONSUMER-ID
test.topic 54 4304 4309 35
consumer-26-21f5050c-a43c-4254-bfcf-42e17dbdb651
test.topic 40 4426 4436 10
consumer-21-24f3ebca-004f-4aac-a348-638c9c6a02f0
test.topic 59 4414 4420 63
consumer-27-ed34f1b3-1be9-422b-bb07-e3c9913195c7
test.topic 42 4389 4403 76
consumer-22-75c2fc0a-5d5c-472d-b27e-e873030f82b6
test.topic 27 4416 4422 24
consumer-18-3be20568-8dd3-4679-a008-0ca64d31083c
On Fri, Jul 7, 2017 at 2:52 PM, M. Manna <ma...@gmail.com> wrote:
> kafka-consumer-groups.sh --bootstrap-server broker:9092 --new-consumer
> --group service-group --describe
>
> how many brokers do you have in the cluster? if you have more than one,
> list them all using a comma csv with --bootstrap-server.
>
> Also, could you paste some results from the console printout?
>
> On 7 July 2017 at 12:47, Dmitriy Vsekhvalnov <dv...@gmail.com>
> wrote:
>
> > Hi all,
> >
> > question about lag checking. We've tried to periodically sample consumer
> > lag with:
> >
> > kafka-consumer-groups.sh --bootstrap-server broker:9092 --new-consumer
> > --group service-group --describe
> >
> > it's all fine, but depending on host we run it from it gives different
> > results.
> >
> > E.g:
> >
> > - when running from one of the broker hosts itself we getting close to
> 0
> > figures.
> >
> > - when running from remote host, we getting 30-60 in average (i suspect
> > there are multiple remote calls to broker involved, so difference due to
> > timing).
> >
> >
> > My question is what is correct way to use it? From broker host itself?
> >
>
Re: Where to run kafka-consumer-groups.sh from?
Posted by "M. Manna" <ma...@gmail.com>.
kafka-consumer-groups.sh --bootstrap-server broker:9092 --new-consumer
--group service-group --describe
how many brokers do you have in the cluster? if you have more than one,
list them all using a comma csv with --bootstrap-server.
Also, could you paste some results from the console printout?
On 7 July 2017 at 12:47, Dmitriy Vsekhvalnov <dv...@gmail.com> wrote:
> Hi all,
>
> question about lag checking. We've tried to periodically sample consumer
> lag with:
>
> kafka-consumer-groups.sh --bootstrap-server broker:9092 --new-consumer
> --group service-group --describe
>
> it's all fine, but depending on host we run it from it gives different
> results.
>
> E.g:
>
> - when running from one of the broker hosts itself we getting close to 0
> figures.
>
> - when running from remote host, we getting 30-60 in average (i suspect
> there are multiple remote calls to broker involved, so difference due to
> timing).
>
>
> My question is what is correct way to use it? From broker host itself?
>