Quantcast

[infinispan-dev] Changing default Hot Rod client max retries

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[infinispan-dev] Changing default Hot Rod client max retries

Galder Zamarreño
Hi all,

Java Hot Rod client has 10 max retries as default. This sounds a bit too much, and as I find the need to add similar configuration to JS client, I'm wondering whether this should be reduce to 3 for all clients, including Java, C* and JS clients.

Any objections?

Cheers,
--
Galder Zamarreño
Infinispan, Red Hat


_______________________________________________
infinispan-dev mailing list
[hidden email]
https://lists.jboss.org/mailman/listinfo/infinispan-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: [infinispan-dev] Changing default Hot Rod client max retries

Sanne Grinovero-3
No objection, just not sure about the usefulness. I think what matters
for people is how long is it going to wait before it fails.

If it's a long time (i.e. 10 minutes) then you'd probably want it try
faster than waiting 5 minutes for the second try ... exponential
backoff sounds nicer than trying to find a reasonable balance in the
connection retries.

Another benefit of an exponential backoff strategy is that you could
allow the users to set an option to wait essentially forever (until
interrupted: nicer to allow this control to higher up stacks), which
could be useful for cloud deployments, microservices, etc..



On 1 June 2016 at 09:26, Galder Zamarreño <[hidden email]> wrote:

> Hi all,
>
> Java Hot Rod client has 10 max retries as default. This sounds a bit too much, and as I find the need to add similar configuration to JS client, I'm wondering whether this should be reduce to 3 for all clients, including Java, C* and JS clients.
>
> Any objections?
>
> Cheers,
> --
> Galder Zamarreño
> Infinispan, Red Hat
>
>
> _______________________________________________
> infinispan-dev mailing list
> [hidden email]
> https://lists.jboss.org/mailman/listinfo/infinispan-dev

_______________________________________________
infinispan-dev mailing list
[hidden email]
https://lists.jboss.org/mailman/listinfo/infinispan-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: [infinispan-dev] Changing default Hot Rod client max retries

Dan Berindei
I'd also like to see an option for the total time to wait, instead of
having to worry about two (or more) different settings.

True, if there's a bug that causes the request to fail immediately and
the client retries without pause for 1 minute, it can generate a lot
of unnecessary load. So perhaps we should only retry if we "know" the
error can be fixed by retrying, e.g. on connection close or on
IllegalLifecycleStateExceptions.

Cheers
Dan


On Wed, Jun 1, 2016 at 12:34 PM, Sanne Grinovero <[hidden email]> wrote:

> No objection, just not sure about the usefulness. I think what matters
> for people is how long is it going to wait before it fails.
>
> If it's a long time (i.e. 10 minutes) then you'd probably want it try
> faster than waiting 5 minutes for the second try ... exponential
> backoff sounds nicer than trying to find a reasonable balance in the
> connection retries.
>
> Another benefit of an exponential backoff strategy is that you could
> allow the users to set an option to wait essentially forever (until
> interrupted: nicer to allow this control to higher up stacks), which
> could be useful for cloud deployments, microservices, etc..
>
>
>
> On 1 June 2016 at 09:26, Galder Zamarreño <[hidden email]> wrote:
>> Hi all,
>>
>> Java Hot Rod client has 10 max retries as default. This sounds a bit too much, and as I find the need to add similar configuration to JS client, I'm wondering whether this should be reduce to 3 for all clients, including Java, C* and JS clients.
>>
>> Any objections?
>>
>> Cheers,
>> --
>> Galder Zamarreño
>> Infinispan, Red Hat
>>
>>
>> _______________________________________________
>> infinispan-dev mailing list
>> [hidden email]
>> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>
> _______________________________________________
> infinispan-dev mailing list
> [hidden email]
> https://lists.jboss.org/mailman/listinfo/infinispan-dev

_______________________________________________
infinispan-dev mailing list
[hidden email]
https://lists.jboss.org/mailman/listinfo/infinispan-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: [infinispan-dev] Changing default Hot Rod client max retries

William Burns-3


On Wed, Jun 1, 2016 at 8:35 AM Dan Berindei <[hidden email]> wrote:
I'd also like to see an option for the total time to wait, instead of
having to worry about two (or more) different settings.

Only 1 config sounds good to me.  I admit I am more used to total time to wait rather than retry, using long and TimeUnit.
 

True, if there's a bug that causes the request to fail immediately and
the client retries without pause for 1 minute, it can generate a lot
of unnecessary load. So perhaps we should only retry if we "know" the
error can be fixed by retrying, e.g. on connection close or on
IllegalLifecycleStateExceptions.

+1, retrying on specific exceptions sounds like a good idea to me
 

Cheers
Dan


On Wed, Jun 1, 2016 at 12:34 PM, Sanne Grinovero <[hidden email]> wrote:
> No objection, just not sure about the usefulness. I think what matters
> for people is how long is it going to wait before it fails.
>
> If it's a long time (i.e. 10 minutes) then you'd probably want it try
> faster than waiting 5 minutes for the second try ... exponential
> backoff sounds nicer than trying to find a reasonable balance in the
> connection retries.
>
> Another benefit of an exponential backoff strategy is that you could
> allow the users to set an option to wait essentially forever (until
> interrupted: nicer to allow this control to higher up stacks), which
> could be useful for cloud deployments, microservices, etc..
>
>
>
> On 1 June 2016 at 09:26, Galder Zamarreño <[hidden email]> wrote:
>> Hi all,
>>
>> Java Hot Rod client has 10 max retries as default. This sounds a bit too much, and as I find the need to add similar configuration to JS client, I'm wondering whether this should be reduce to 3 for all clients, including Java, C* and JS clients.
>>
>> Any objections?
>>
>> Cheers,
>> --
>> Galder Zamarreño
>> Infinispan, Red Hat
>>
>>
>> _______________________________________________
>> infinispan-dev mailing list
>> [hidden email]
>> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>
> _______________________________________________
> infinispan-dev mailing list
> [hidden email]
> https://lists.jboss.org/mailman/listinfo/infinispan-dev

_______________________________________________
infinispan-dev mailing list
[hidden email]
https://lists.jboss.org/mailman/listinfo/infinispan-dev

_______________________________________________
infinispan-dev mailing list
[hidden email]
https://lists.jboss.org/mailman/listinfo/infinispan-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: [infinispan-dev] Changing default Hot Rod client max retries

Galder Zamarreño
Thanks all for the feedback. Very good points made, so I've created this: ISPN-6774 and HRJS-22 so that we implement that for Java/JS clients.

@Vittorio, how does C++/C# client deal with this?

Cheers,
--
Galder Zamarreño
Infinispan, Red Hat

> On 2 Jun 2016, at 19:38, William Burns <[hidden email]> wrote:
>
>
>
> On Wed, Jun 1, 2016 at 8:35 AM Dan Berindei <[hidden email]> wrote:
> I'd also like to see an option for the total time to wait, instead of
> having to worry about two (or more) different settings.
>
> Only 1 config sounds good to me.  I admit I am more used to total time to wait rather than retry, using long and TimeUnit.
>  
>
> True, if there's a bug that causes the request to fail immediately and
> the client retries without pause for 1 minute, it can generate a lot
> of unnecessary load. So perhaps we should only retry if we "know" the
> error can be fixed by retrying, e.g. on connection close or on
> IllegalLifecycleStateExceptions.
>
> +1, retrying on specific exceptions sounds like a good idea to me
>  
>
> Cheers
> Dan
>
>
> On Wed, Jun 1, 2016 at 12:34 PM, Sanne Grinovero <[hidden email]> wrote:
> > No objection, just not sure about the usefulness. I think what matters
> > for people is how long is it going to wait before it fails.
> >
> > If it's a long time (i.e. 10 minutes) then you'd probably want it try
> > faster than waiting 5 minutes for the second try ... exponential
> > backoff sounds nicer than trying to find a reasonable balance in the
> > connection retries.
> >
> > Another benefit of an exponential backoff strategy is that you could
> > allow the users to set an option to wait essentially forever (until
> > interrupted: nicer to allow this control to higher up stacks), which
> > could be useful for cloud deployments, microservices, etc..
> >
> >
> >
> > On 1 June 2016 at 09:26, Galder Zamarreño <[hidden email]> wrote:
> >> Hi all,
> >>
> >> Java Hot Rod client has 10 max retries as default. This sounds a bit too much, and as I find the need to add similar configuration to JS client, I'm wondering whether this should be reduce to 3 for all clients, including Java, C* and JS clients.
> >>
> >> Any objections?
> >>
> >> Cheers,
> >> --
> >> Galder Zamarreño
> >> Infinispan, Red Hat
> >>
> >>
> >> _______________________________________________
> >> infinispan-dev mailing list
> >> [hidden email]
> >> https://lists.jboss.org/mailman/listinfo/infinispan-dev
> >
> > _______________________________________________
> > infinispan-dev mailing list
> > [hidden email]
> > https://lists.jboss.org/mailman/listinfo/infinispan-dev
>
> _______________________________________________
> infinispan-dev mailing list
> [hidden email]
> https://lists.jboss.org/mailman/listinfo/infinispan-dev
> _______________________________________________
> infinispan-dev mailing list
> [hidden email]
> https://lists.jboss.org/mailman/listinfo/infinispan-dev


_______________________________________________
infinispan-dev mailing list
[hidden email]
https://lists.jboss.org/mailman/listinfo/infinispan-dev
Loading...