You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hc.apache.org by "Dan Checkoway (JIRA)" <ji...@apache.org> on 2011/06/24 14:05:47 UTC
[jira] [Created] (HTTPCLIENT-1105) Built-in way to do auto-retry
for certain status codes
Built-in way to do auto-retry for certain status codes
------------------------------------------------------
Key: HTTPCLIENT-1105
URL: https://issues.apache.org/jira/browse/HTTPCLIENT-1105
Project: HttpComponents HttpClient
Issue Type: Wish
Components: HttpClient
Affects Versions: 4.1.1
Reporter: Dan Checkoway
Priority: Minor
The HttpRequestRetryHandler mechanism is great. It allows API users to plug in their own logic to control whether or not a retry should automatically be done, how many times it should be retried, etc. That works perfectly in scenarios where an exception is caught while issuing the request. It falls short, however, in this scenario...
If I'm hitting a service that returns a 503, I want to be able to retry that request automatically as well. As of right now, I need to write my own logic to accomplish that, and it's clunky trying to integrate it with the httpClient.execute() call, since it's my ResponseHandler impl that ends up getting the 503. I can see use cases for auto-retrying upon getting other HTTP statuses as well, not just 503.
My request here is...I would love to be able to configure, either on the HttpClient itself or on a wrapping class or something, that a request should automatically be retried if the HTTP status code is among of a set of statuses that I configure. It would be nice if you could set the max # of retries, an optional sleep time in between retries (perhaps optional incremental backoff if you want to get fancy). I'm not sure if this is possible, but it would be nice if -- when this type of status-based retry is enabled -- my ResponseHandler wouldn't even get invoked until retry was successful.
Here's an alternative suggestion, possibly simpler to build, but definitely not as elegant:
In my ResponseHandler, you could throw a RetryRequestException or something like that, and the calling code would catch that and do as expected. That might simplify the mechanism so to speak.
Anyway, I would love not to have to roll my own retry code, since I suspect this is something that hundreds (thousands?) of HttpClient users have had to code. Seems like providing a standardized, well-written way to do it would go a long way to helping many coders out there.
Thanks!
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org
[jira] [Updated] (HTTPCLIENT-1105) Built-in way to do auto-retry
for certain status codes
Posted by "Oleg Kalnichevski (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HTTPCLIENT-1105?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Oleg Kalnichevski updated HTTPCLIENT-1105:
------------------------------------------
Fix Version/s: Future
Probably the best and cleanest way to solve the problem would be a decorator class for HttpClient interface similar to CachingHttpClient. It should be fairly easy to implement.
We happily take contributions.
Oleg
> Built-in way to do auto-retry for certain status codes
> ------------------------------------------------------
>
> Key: HTTPCLIENT-1105
> URL: https://issues.apache.org/jira/browse/HTTPCLIENT-1105
> Project: HttpComponents HttpClient
> Issue Type: Wish
> Components: HttpClient
> Affects Versions: 4.1.1
> Reporter: Dan Checkoway
> Priority: Minor
> Fix For: Future
>
>
> The HttpRequestRetryHandler mechanism is great. It allows API users to plug in their own logic to control whether or not a retry should automatically be done, how many times it should be retried, etc. That works perfectly in scenarios where an exception is caught while issuing the request. It falls short, however, in this scenario...
> If I'm hitting a service that returns a 503, I want to be able to retry that request automatically as well. As of right now, I need to write my own logic to accomplish that, and it's clunky trying to integrate it with the httpClient.execute() call, since it's my ResponseHandler impl that ends up getting the 503. I can see use cases for auto-retrying upon getting other HTTP statuses as well, not just 503.
> My request here is...I would love to be able to configure, either on the HttpClient itself or on a wrapping class or something, that a request should automatically be retried if the HTTP status code is among of a set of statuses that I configure. It would be nice if you could set the max # of retries, an optional sleep time in between retries (perhaps optional incremental backoff if you want to get fancy). I'm not sure if this is possible, but it would be nice if -- when this type of status-based retry is enabled -- my ResponseHandler wouldn't even get invoked until retry was successful.
> Here's an alternative suggestion, possibly simpler to build, but definitely not as elegant:
> In my ResponseHandler, you could throw a RetryRequestException or something like that, and the calling code would catch that and do as expected. That might simplify the mechanism so to speak.
> Anyway, I would love not to have to roll my own retry code, since I suspect this is something that hundreds (thousands?) of HttpClient users have had to code. Seems like providing a standardized, well-written way to do it would go a long way to helping many coders out there.
> Thanks!
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org
[jira] [Updated] (HTTPCLIENT-1105) Built-in way to do auto-retry
for certain status codes
Posted by "Alin Vasile (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HTTPCLIENT-1105?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alin Vasile updated HTTPCLIENT-1105:
------------------------------------
Attachment: patch_auto_retry_v2.diff
Good idea. I attached the updated path with you comment, plus some javadocs and retry functionality updated.
> Built-in way to do auto-retry for certain status codes
> ------------------------------------------------------
>
> Key: HTTPCLIENT-1105
> URL: https://issues.apache.org/jira/browse/HTTPCLIENT-1105
> Project: HttpComponents HttpClient
> Issue Type: Wish
> Components: HttpClient
> Affects Versions: 4.1.1
> Reporter: Dan Checkoway
> Priority: Minor
> Fix For: 4.2 Alpha1
>
> Attachments: patch_auto_retry.diff, patch_auto_retry_v2.diff
>
>
> The HttpRequestRetryHandler mechanism is great. It allows API users to plug in their own logic to control whether or not a retry should automatically be done, how many times it should be retried, etc. That works perfectly in scenarios where an exception is caught while issuing the request. It falls short, however, in this scenario...
> If I'm hitting a service that returns a 503, I want to be able to retry that request automatically as well. As of right now, I need to write my own logic to accomplish that, and it's clunky trying to integrate it with the httpClient.execute() call, since it's my ResponseHandler impl that ends up getting the 503. I can see use cases for auto-retrying upon getting other HTTP statuses as well, not just 503.
> My request here is...I would love to be able to configure, either on the HttpClient itself or on a wrapping class or something, that a request should automatically be retried if the HTTP status code is among of a set of statuses that I configure. It would be nice if you could set the max # of retries, an optional sleep time in between retries (perhaps optional incremental backoff if you want to get fancy). I'm not sure if this is possible, but it would be nice if -- when this type of status-based retry is enabled -- my ResponseHandler wouldn't even get invoked until retry was successful.
> Here's an alternative suggestion, possibly simpler to build, but definitely not as elegant:
> In my ResponseHandler, you could throw a RetryRequestException or something like that, and the calling code would catch that and do as expected. That might simplify the mechanism so to speak.
> Anyway, I would love not to have to roll my own retry code, since I suspect this is something that hundreds (thousands?) of HttpClient users have had to code. Seems like providing a standardized, well-written way to do it would go a long way to helping many coders out there.
> Thanks!
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org
[jira] [Updated] (HTTPCLIENT-1105) Built-in way to do auto-retry
for certain status codes
Posted by "Alin Vasile (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HTTPCLIENT-1105?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alin Vasile updated HTTPCLIENT-1105:
------------------------------------
Attachment: patch_auto_retry.diff
Sample work for the above functionality. Most of the work is copied from the CachingHttpClient. My environment is not fully functional, so I wasn't able to optimize the imports. Please share your thoughts...
> Built-in way to do auto-retry for certain status codes
> ------------------------------------------------------
>
> Key: HTTPCLIENT-1105
> URL: https://issues.apache.org/jira/browse/HTTPCLIENT-1105
> Project: HttpComponents HttpClient
> Issue Type: Wish
> Components: HttpClient
> Affects Versions: 4.1.1
> Reporter: Dan Checkoway
> Priority: Minor
> Fix For: Future
>
> Attachments: patch_auto_retry.diff
>
>
> The HttpRequestRetryHandler mechanism is great. It allows API users to plug in their own logic to control whether or not a retry should automatically be done, how many times it should be retried, etc. That works perfectly in scenarios where an exception is caught while issuing the request. It falls short, however, in this scenario...
> If I'm hitting a service that returns a 503, I want to be able to retry that request automatically as well. As of right now, I need to write my own logic to accomplish that, and it's clunky trying to integrate it with the httpClient.execute() call, since it's my ResponseHandler impl that ends up getting the 503. I can see use cases for auto-retrying upon getting other HTTP statuses as well, not just 503.
> My request here is...I would love to be able to configure, either on the HttpClient itself or on a wrapping class or something, that a request should automatically be retried if the HTTP status code is among of a set of statuses that I configure. It would be nice if you could set the max # of retries, an optional sleep time in between retries (perhaps optional incremental backoff if you want to get fancy). I'm not sure if this is possible, but it would be nice if -- when this type of status-based retry is enabled -- my ResponseHandler wouldn't even get invoked until retry was successful.
> Here's an alternative suggestion, possibly simpler to build, but definitely not as elegant:
> In my ResponseHandler, you could throw a RetryRequestException or something like that, and the calling code would catch that and do as expected. That might simplify the mechanism so to speak.
> Anyway, I would love not to have to roll my own retry code, since I suspect this is something that hundreds (thousands?) of HttpClient users have had to code. Seems like providing a standardized, well-written way to do it would go a long way to helping many coders out there.
> Thanks!
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org
[jira] [Resolved] (HTTPCLIENT-1105) Built-in way to do auto-retry
for certain status codes
Posted by "Oleg Kalnichevski (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HTTPCLIENT-1105?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Oleg Kalnichevski resolved HTTPCLIENT-1105.
-------------------------------------------
Resolution: Fixed
Patch committed with some changes. I changed the signature of the ServiceUnavailableRetryStrategy to make it more consistent with HttpRequestRetryHandler. I also did not quite like recursive invocation of the #callBackend method. I replaced recursive method call with a simple loop.
Please review.
I also noticed that the default ServiceUnavailableRetryStrategy implementation was not fully thread safe. Ideally all instance variables in the DefaultServiceUnavailableRetryStrategy should be made final and the class should be made immutable.
Otherwise, many thanks for the contribution.
Oleg
> Built-in way to do auto-retry for certain status codes
> ------------------------------------------------------
>
> Key: HTTPCLIENT-1105
> URL: https://issues.apache.org/jira/browse/HTTPCLIENT-1105
> Project: HttpComponents HttpClient
> Issue Type: Wish
> Components: HttpClient
> Affects Versions: 4.1.1
> Reporter: Dan Checkoway
> Priority: Minor
> Fix For: 4.2 Alpha1
>
> Attachments: patch_auto_retry.diff, patch_auto_retry_v2.diff
>
>
> The HttpRequestRetryHandler mechanism is great. It allows API users to plug in their own logic to control whether or not a retry should automatically be done, how many times it should be retried, etc. That works perfectly in scenarios where an exception is caught while issuing the request. It falls short, however, in this scenario...
> If I'm hitting a service that returns a 503, I want to be able to retry that request automatically as well. As of right now, I need to write my own logic to accomplish that, and it's clunky trying to integrate it with the httpClient.execute() call, since it's my ResponseHandler impl that ends up getting the 503. I can see use cases for auto-retrying upon getting other HTTP statuses as well, not just 503.
> My request here is...I would love to be able to configure, either on the HttpClient itself or on a wrapping class or something, that a request should automatically be retried if the HTTP status code is among of a set of statuses that I configure. It would be nice if you could set the max # of retries, an optional sleep time in between retries (perhaps optional incremental backoff if you want to get fancy). I'm not sure if this is possible, but it would be nice if -- when this type of status-based retry is enabled -- my ResponseHandler wouldn't even get invoked until retry was successful.
> Here's an alternative suggestion, possibly simpler to build, but definitely not as elegant:
> In my ResponseHandler, you could throw a RetryRequestException or something like that, and the calling code would catch that and do as expected. That might simplify the mechanism so to speak.
> Anyway, I would love not to have to roll my own retry code, since I suspect this is something that hundreds (thousands?) of HttpClient users have had to code. Seems like providing a standardized, well-written way to do it would go a long way to helping many coders out there.
> Thanks!
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org
[jira] [Updated] (HTTPCLIENT-1105) Built-in way to do auto-retry
for certain status codes
Posted by "Oleg Kalnichevski (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HTTPCLIENT-1105?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Oleg Kalnichevski updated HTTPCLIENT-1105:
------------------------------------------
Fix Version/s: (was: Future)
4.2 Alpha1
Looks reasonable to me. I just would make an abstract strategy interface out of ServiceUnavailableRetryConfig.
Oleg
> Built-in way to do auto-retry for certain status codes
> ------------------------------------------------------
>
> Key: HTTPCLIENT-1105
> URL: https://issues.apache.org/jira/browse/HTTPCLIENT-1105
> Project: HttpComponents HttpClient
> Issue Type: Wish
> Components: HttpClient
> Affects Versions: 4.1.1
> Reporter: Dan Checkoway
> Priority: Minor
> Fix For: 4.2 Alpha1
>
> Attachments: patch_auto_retry.diff
>
>
> The HttpRequestRetryHandler mechanism is great. It allows API users to plug in their own logic to control whether or not a retry should automatically be done, how many times it should be retried, etc. That works perfectly in scenarios where an exception is caught while issuing the request. It falls short, however, in this scenario...
> If I'm hitting a service that returns a 503, I want to be able to retry that request automatically as well. As of right now, I need to write my own logic to accomplish that, and it's clunky trying to integrate it with the httpClient.execute() call, since it's my ResponseHandler impl that ends up getting the 503. I can see use cases for auto-retrying upon getting other HTTP statuses as well, not just 503.
> My request here is...I would love to be able to configure, either on the HttpClient itself or on a wrapping class or something, that a request should automatically be retried if the HTTP status code is among of a set of statuses that I configure. It would be nice if you could set the max # of retries, an optional sleep time in between retries (perhaps optional incremental backoff if you want to get fancy). I'm not sure if this is possible, but it would be nice if -- when this type of status-based retry is enabled -- my ResponseHandler wouldn't even get invoked until retry was successful.
> Here's an alternative suggestion, possibly simpler to build, but definitely not as elegant:
> In my ResponseHandler, you could throw a RetryRequestException or something like that, and the calling code would catch that and do as expected. That might simplify the mechanism so to speak.
> Anyway, I would love not to have to roll my own retry code, since I suspect this is something that hundreds (thousands?) of HttpClient users have had to code. Seems like providing a standardized, well-written way to do it would go a long way to helping many coders out there.
> Thanks!
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org
[jira] [Commented] (HTTPCLIENT-1105) Built-in way to do auto-retry
for certain status codes
Posted by "Alin Vasile (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HTTPCLIENT-1105?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13057459#comment-13057459 ]
Alin Vasile commented on HTTPCLIENT-1105:
-----------------------------------------
Looks OK to me...
> Built-in way to do auto-retry for certain status codes
> ------------------------------------------------------
>
> Key: HTTPCLIENT-1105
> URL: https://issues.apache.org/jira/browse/HTTPCLIENT-1105
> Project: HttpComponents HttpClient
> Issue Type: Wish
> Components: HttpClient
> Affects Versions: 4.1.1
> Reporter: Dan Checkoway
> Priority: Minor
> Fix For: 4.2 Alpha1
>
> Attachments: patch_auto_retry.diff, patch_auto_retry_v2.diff
>
>
> The HttpRequestRetryHandler mechanism is great. It allows API users to plug in their own logic to control whether or not a retry should automatically be done, how many times it should be retried, etc. That works perfectly in scenarios where an exception is caught while issuing the request. It falls short, however, in this scenario...
> If I'm hitting a service that returns a 503, I want to be able to retry that request automatically as well. As of right now, I need to write my own logic to accomplish that, and it's clunky trying to integrate it with the httpClient.execute() call, since it's my ResponseHandler impl that ends up getting the 503. I can see use cases for auto-retrying upon getting other HTTP statuses as well, not just 503.
> My request here is...I would love to be able to configure, either on the HttpClient itself or on a wrapping class or something, that a request should automatically be retried if the HTTP status code is among of a set of statuses that I configure. It would be nice if you could set the max # of retries, an optional sleep time in between retries (perhaps optional incremental backoff if you want to get fancy). I'm not sure if this is possible, but it would be nice if -- when this type of status-based retry is enabled -- my ResponseHandler wouldn't even get invoked until retry was successful.
> Here's an alternative suggestion, possibly simpler to build, but definitely not as elegant:
> In my ResponseHandler, you could throw a RetryRequestException or something like that, and the calling code would catch that and do as expected. That might simplify the mechanism so to speak.
> Anyway, I would love not to have to roll my own retry code, since I suspect this is something that hundreds (thousands?) of HttpClient users have had to code. Seems like providing a standardized, well-written way to do it would go a long way to helping many coders out there.
> Thanks!
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org