You are viewing a plain text version of this content. The canonical link for it is here.
Posted to httpclient-users@hc.apache.org by "Alex, Joseph (Contractor)" <Jo...@searshc.com> on 2014/03/03 13:59:53 UTC
Connection Pool and DNS Resolution timeout
Hi,
We use HttpClient 3.1 and Httpcomponents 4.2.x in some of our apps with MultiThreadedHttpConnectionManager and PoolingClientConnectionManager respectively. Some of our server endpoints use a Global Name which can be switched to point to different data centers at runtime, depending on which one is active (with the other DC still being "alive"). Example www.global.com<http://www.global.com> which could either point to www.dc1.com<http://www.dc1.com> or www.dc2.com<http://www.dc2.com>. When this switch was done, we noticed that the connections were not picking up the new DC's IP and still routing to the old one. Assuming that the DNS resolution timeouts (networkaddress.cache.ttl) is set to a low value in the JVM, would the connection pool honor this value and reset the connections accordingly? . If they don't, what options do we have of ensuring that the connections are reset, and recreated with the new DNS resolution, without having to restart the JVM?
Thanks,
Joseph
This message, including any attachments, is the property of Sears Holdings Corporation and/or one of its subsidiaries. It is confidential and may contain proprietary or legally privileged information. If you are not the intended recipient, please delete it without reading the contents. Thank you.
Re: Connection Pool and DNS Resolution timeout
Posted by Oleg Kalnichevski <ol...@apache.org>.
On Thu, 2014-03-06 at 13:12 +0000, Alex, Joseph (Contractor) wrote:
> Thanks Oleg. This will help for new connections getting created to pick up the new IP. But, how can we manage connections that were established and exist in the pool when the DC switch is made? . I saw that connectionmanger shutdown() will close all connections, but that would also make it and the HttpClients unusable unless they are initialized again. Is there any way to just flush all existing connections?
>
HttpClientConnectionManager#closeIdleConnections(0, TimeUnit.SECONDS) is
your friend.
http://hc.apache.org/httpcomponents-client-4.3.x/tutorial/html/connmgmt.html#d5e380
Oleg
> -Joseph
>
> -----Original Message-----
> From: Oleg Kalnichevski [mailto:olegk@apache.org]
> Sent: Tuesday, March 04, 2014 8:14 AM
> To: HttpClient User Discussion
> Subject: Re: Connection Pool and DNS Resolution timeout
>
> On Mon, 2014-03-03 at 12:59 +0000, Alex, Joseph (Contractor) wrote:
> > Hi,
> >
> > We use HttpClient 3.1 and Httpcomponents 4.2.x in some of our apps with MultiThreadedHttpConnectionManager and PoolingClientConnectionManager respectively. Some of our server endpoints use a Global Name which can be switched to point to different data centers at runtime, depending on which one is active (with the other DC still being "alive"). Example www.global.com<http://www.global.com> which could either point to www.dc1.com<http://www.dc1.com> or www.dc2.com<http://www.dc2.com>. When this switch was done, we noticed that the connections were not picking up the new DC's IP and still routing to the old one. Assuming that the DNS resolution timeouts (networkaddress.cache.ttl) is set to a low value in the JVM, would the connection pool honor this value and reset the connections accordingly? . If they don't, what options do we have of ensuring that the connections are reset, and recreated with the new DNS resolution, without having to restart the JVM?
> >
> > Thanks,
> > Joseph
>
> Hi Joseph
>
> HttpClient of all versions are fully reliant on the standard DNS resolution mechanism provided by JRE. They also do not attempt to cache or re-use resolved addresses, so 'networkaddress.cache.ttl' parameter should have effect.
>
> Since HttpClient 4.2 though, one can plug in a custom DNS resolver implementation [1] to replace or enhance the standard resolution mechanism.
>
> Hope this helps
>
> Oleg
>
> [1]
> http://hc.apache.org/httpcomponents-client-4.3.x/httpclient/apidocs/org/apache/http/conn/DnsResolver.html
>
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: httpclient-users-unsubscribe@hc.apache.org
> For additional commands, e-mail: httpclient-users-help@hc.apache.org
>
>
> This message, including any attachments, is the property of Sears Holdings Corporation and/or one of its subsidiaries. It is confidential and may contain proprietary or legally privileged information. If you are not the intended recipient, please delete it without reading the contents. Thank you.
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: httpclient-users-unsubscribe@hc.apache.org
> For additional commands, e-mail: httpclient-users-help@hc.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: httpclient-users-unsubscribe@hc.apache.org
For additional commands, e-mail: httpclient-users-help@hc.apache.org
RE: Connection Pool and DNS Resolution timeout
Posted by "Alex, Joseph (Contractor)" <Jo...@searshc.com>.
Thanks Oleg. This will help for new connections getting created to pick up the new IP. But, how can we manage connections that were established and exist in the pool when the DC switch is made? . I saw that connectionmanger shutdown() will close all connections, but that would also make it and the HttpClients unusable unless they are initialized again. Is there any way to just flush all existing connections?
-Joseph
-----Original Message-----
From: Oleg Kalnichevski [mailto:olegk@apache.org]
Sent: Tuesday, March 04, 2014 8:14 AM
To: HttpClient User Discussion
Subject: Re: Connection Pool and DNS Resolution timeout
On Mon, 2014-03-03 at 12:59 +0000, Alex, Joseph (Contractor) wrote:
> Hi,
>
> We use HttpClient 3.1 and Httpcomponents 4.2.x in some of our apps with MultiThreadedHttpConnectionManager and PoolingClientConnectionManager respectively. Some of our server endpoints use a Global Name which can be switched to point to different data centers at runtime, depending on which one is active (with the other DC still being "alive"). Example www.global.com<http://www.global.com> which could either point to www.dc1.com<http://www.dc1.com> or www.dc2.com<http://www.dc2.com>. When this switch was done, we noticed that the connections were not picking up the new DC's IP and still routing to the old one. Assuming that the DNS resolution timeouts (networkaddress.cache.ttl) is set to a low value in the JVM, would the connection pool honor this value and reset the connections accordingly? . If they don't, what options do we have of ensuring that the connections are reset, and recreated with the new DNS resolution, without having to restart the JVM?
>
> Thanks,
> Joseph
Hi Joseph
HttpClient of all versions are fully reliant on the standard DNS resolution mechanism provided by JRE. They also do not attempt to cache or re-use resolved addresses, so 'networkaddress.cache.ttl' parameter should have effect.
Since HttpClient 4.2 though, one can plug in a custom DNS resolver implementation [1] to replace or enhance the standard resolution mechanism.
Hope this helps
Oleg
[1]
http://hc.apache.org/httpcomponents-client-4.3.x/httpclient/apidocs/org/apache/http/conn/DnsResolver.html
---------------------------------------------------------------------
To unsubscribe, e-mail: httpclient-users-unsubscribe@hc.apache.org
For additional commands, e-mail: httpclient-users-help@hc.apache.org
This message, including any attachments, is the property of Sears Holdings Corporation and/or one of its subsidiaries. It is confidential and may contain proprietary or legally privileged information. If you are not the intended recipient, please delete it without reading the contents. Thank you.
Re: Connection Pool and DNS Resolution timeout
Posted by Oleg Kalnichevski <ol...@apache.org>.
On Mon, 2014-03-03 at 12:59 +0000, Alex, Joseph (Contractor) wrote:
> Hi,
>
> We use HttpClient 3.1 and Httpcomponents 4.2.x in some of our apps with MultiThreadedHttpConnectionManager and PoolingClientConnectionManager respectively. Some of our server endpoints use a Global Name which can be switched to point to different data centers at runtime, depending on which one is active (with the other DC still being "alive"). Example www.global.com<http://www.global.com> which could either point to www.dc1.com<http://www.dc1.com> or www.dc2.com<http://www.dc2.com>. When this switch was done, we noticed that the connections were not picking up the new DC's IP and still routing to the old one. Assuming that the DNS resolution timeouts (networkaddress.cache.ttl) is set to a low value in the JVM, would the connection pool honor this value and reset the connections accordingly? . If they don't, what options do we have of ensuring that the connections are reset, and recreated with the new DNS resolution, without having to restart the JVM?
>
> Thanks,
> Joseph
Hi Joseph
HttpClient of all versions are fully reliant on the standard DNS
resolution mechanism provided by JRE. They also do not attempt to cache
or re-use resolved addresses, so 'networkaddress.cache.ttl' parameter
should have effect.
Since HttpClient 4.2 though, one can plug in a custom DNS resolver
implementation [1] to replace or enhance the standard resolution
mechanism.
Hope this helps
Oleg
[1]
http://hc.apache.org/httpcomponents-client-4.3.x/httpclient/apidocs/org/apache/http/conn/DnsResolver.html
---------------------------------------------------------------------
To unsubscribe, e-mail: httpclient-users-unsubscribe@hc.apache.org
For additional commands, e-mail: httpclient-users-help@hc.apache.org