You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@trafficserver.apache.org by "Leif Hedstrom (JIRA)" <ji...@apache.org> on 2010/11/30 19:40:11 UTC
[jira] Created: (TS-566) Rename TCP_DEFER_ACCEPT option and change
default
Rename TCP_DEFER_ACCEPT option and change default
-------------------------------------------------
Key: TS-566
URL: https://issues.apache.org/jira/browse/TS-566
Project: Traffic Server
Issue Type: Improvement
Components: Network
Reporter: Leif Hedstrom
Fix For: 2.1.5
Two proposals:
1) We rename the existing config from proxy.config.net.tcp_accept_defer_timeout to proxy.config.net.defer_accept . On Linux, this option is a timeout value (as used with the TCP_DEFER_ACCEPT sysctl), and on other platforms (e.g. on FreeBSD), it would be a flag to enable / disable deferred accepts / accept filters.
2) We change the default of this setting to 45 on Linux. This would be functionally identical to the HTTPD v2.3.x setting of 30, but 45 makes somewhat more sense since it's close (or identical) to what the effective timeout will be (it's quantized based on how many retransmits TCP will do). On other platforms. we'll leave this value at "1" in the default (generated) records.config.
Comments welcome!
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (TS-566) Rename TCP_DEFER_ACCEPT option and
change default
Posted by "Leif Hedstrom (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/TS-566?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12966206#action_12966206 ]
Leif Hedstrom commented on TS-566:
----------------------------------
As discussed, we'll rename this setting to be more generic:
CONFIG proxy.config.net.defer_accept INT <value>
where <value> is 0 - ... for Linux, and 0-1 on FreeBSD (assuming someone implements it).
I will set the default for to 45 on linux, and 0 for everything else. For an explanation why 45 is a reasonable default, see http://www.ogre.com/content/tcp-socket-options
Further explanation: Apache HTTPD v2.3 sets this to "30", but that really translates to 45 on modern linux platforms, and 48 on older platforms, so 45 seems like a more reasonable default that's more close to the truth.
> Rename TCP_DEFER_ACCEPT option and change default
> -------------------------------------------------
>
> Key: TS-566
> URL: https://issues.apache.org/jira/browse/TS-566
> Project: Traffic Server
> Issue Type: Improvement
> Components: Network
> Reporter: Leif Hedstrom
> Assignee: Leif Hedstrom
> Fix For: 2.1.5
>
>
> Two proposals:
> 1) We rename the existing config from proxy.config.net.tcp_accept_defer_timeout to proxy.config.net.defer_accept . On Linux, this option is a timeout value (as used with the TCP_DEFER_ACCEPT sysctl), and on other platforms (e.g. on FreeBSD), it would be a flag to enable / disable deferred accepts / accept filters.
> 2) We change the default of this setting to 45 on Linux. This would be functionally identical to the HTTPD v2.3.x setting of 30, but 45 makes somewhat more sense since it's close (or identical) to what the effective timeout will be (it's quantized based on how many retransmits TCP will do). On other platforms. we'll leave this value at "1" in the default (generated) records.config.
> Comments welcome!
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Resolved: (TS-566) Rename TCP_DEFER_ACCEPT option and change
default
Posted by "Leif Hedstrom (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/TS-566?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Leif Hedstrom resolved TS-566.
------------------------------
Resolution: Fixed
> Rename TCP_DEFER_ACCEPT option and change default
> -------------------------------------------------
>
> Key: TS-566
> URL: https://issues.apache.org/jira/browse/TS-566
> Project: Traffic Server
> Issue Type: Improvement
> Components: Network
> Reporter: Leif Hedstrom
> Assignee: Leif Hedstrom
> Fix For: 2.1.5
>
>
> Two proposals:
> 1) We rename the existing config from proxy.config.net.tcp_accept_defer_timeout to proxy.config.net.defer_accept . On Linux, this option is a timeout value (as used with the TCP_DEFER_ACCEPT sysctl), and on other platforms (e.g. on FreeBSD), it would be a flag to enable / disable deferred accepts / accept filters.
> 2) We change the default of this setting to 45 on Linux. This would be functionally identical to the HTTPD v2.3.x setting of 30, but 45 makes somewhat more sense since it's close (or identical) to what the effective timeout will be (it's quantized based on how many retransmits TCP will do). On other platforms. we'll leave this value at "1" in the default (generated) records.config.
> Comments welcome!
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Assigned: (TS-566) Rename TCP_DEFER_ACCEPT option and change
default
Posted by "Leif Hedstrom (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/TS-566?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Leif Hedstrom reassigned TS-566:
--------------------------------
Assignee: Leif Hedstrom
> Rename TCP_DEFER_ACCEPT option and change default
> -------------------------------------------------
>
> Key: TS-566
> URL: https://issues.apache.org/jira/browse/TS-566
> Project: Traffic Server
> Issue Type: Improvement
> Components: Network
> Reporter: Leif Hedstrom
> Assignee: Leif Hedstrom
> Fix For: 2.1.5
>
>
> Two proposals:
> 1) We rename the existing config from proxy.config.net.tcp_accept_defer_timeout to proxy.config.net.defer_accept . On Linux, this option is a timeout value (as used with the TCP_DEFER_ACCEPT sysctl), and on other platforms (e.g. on FreeBSD), it would be a flag to enable / disable deferred accepts / accept filters.
> 2) We change the default of this setting to 45 on Linux. This would be functionally identical to the HTTPD v2.3.x setting of 30, but 45 makes somewhat more sense since it's close (or identical) to what the effective timeout will be (it's quantized based on how many retransmits TCP will do). On other platforms. we'll leave this value at "1" in the default (generated) records.config.
> Comments welcome!
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.