You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@storm.apache.org by Mohan Pandiyan <mo...@pyze.com> on 2015/11/21 02:47:57 UTC
Trident Drpc - Scaling
We are using Drpc - Trident topology. I see my process latency for one of the function (bolt ) is very high compared to the execute latency. This cause whole topologies complete latency to shoot up. I am wondering what would cause the process latency to shoot up 50% more than execute latency and also the capacity is well below .5. Parallelism hint for this level is higher compared all previous level would the repartition cause the process latency to be higher?
PFB screen shot of UI for more info (b1 bolt here along with b6 has higher process latency). Kindly direct me to other thread if this topic has been discussed already.
Regards,
Mohan
Re: Trident Drpc - Scaling
Posted by Mohan Pandiyan <mo...@pyze.com>.
Can anybody help me with this?
Regards,
Mohan
> On Nov 23, 2015, at 4:44 PM, Mohan Pandiyan <mo...@pyze.com> wrote:
>
> I also see that not all the ackers in the topology are receiving acks from bolts (out of 8 only 4 is receiving acks).
>
>
>> On Nov 20, 2015, at 5:47 PM, Mohan Pandiyan <mo...@pyze.com> wrote:
>>
>> We are using Drpc - Trident topology. I see my process latency for one of the function (bolt ) is very high compared to the execute latency. This cause whole topologies complete latency to shoot up. I am wondering what would cause the process latency to shoot up 50% more than execute latency and also the capacity is well below .5. Parallelism hint for this level is higher compared all previous level would the repartition cause the process latency to be higher?
>>
>> PFB screen shot of UI for more info (b1 bolt here along with b6 has higher process latency). Kindly direct me to other thread if this topic has been discussed already.
>>
>> Regards,
>> Mohan
>>
>> <Screen Shot 2015-11-20 at 5.41.57 PM.png>
>
Re: Trident Drpc - Scaling
Posted by Mohan Pandiyan <mo...@pyze.com>.
I also see that not all the ackers in the topology are receiving acks from bolts (out of 8 only 4 is receiving acks).
> On Nov 20, 2015, at 5:47 PM, Mohan Pandiyan <mo...@pyze.com> wrote:
>
> We are using Drpc - Trident topology. I see my process latency for one of the function (bolt ) is very high compared to the execute latency. This cause whole topologies complete latency to shoot up. I am wondering what would cause the process latency to shoot up 50% more than execute latency and also the capacity is well below .5. Parallelism hint for this level is higher compared all previous level would the repartition cause the process latency to be higher?
>
> PFB screen shot of UI for more info (b1 bolt here along with b6 has higher process latency). Kindly direct me to other thread if this topic has been discussed already.
>
> Regards,
> Mohan
>
> <Screen Shot 2015-11-20 at 5.41.57 PM.png>