The issue with changing the configuration value from 60 to 6 is that provides the default value for _METHOD_TIMEOUTS which means now each RPC request will timeout after 6 seconds rather than 60 seconds.  In a heavily loaded system that timeout will get hit frequently thus leading to repeated RPC requests and an increased load on the server. 

 

We need a way to control the maximum timeout without reducing the timeout of each individual RPC timeout so much as to increase the load on the server.

 

I think it is unlikely that upstream will accept the current change in its current form.  Perhaps we need to propose a configuration option that can be controlled via config file but still leaves the default as it is today.

 

Regards,

Allain

 

 

Allain Legacy, Software Developer, Wind River
direct 613.270.2279  fax 613.492.7870 skype allain.legacy

350 Terry Fox Drive, Suite 200, Ottawa, Ontario, K2K 2W5


 
WIND

 

From: Xu, Chenjie [mailto:chenjie.xu@intel.com]
Sent: Tuesday, November 06, 2018 2:35 AM
To: Peters, Matt
Cc: starlingx-discuss@lists.starlingx.io
Subject: [Starlingx-discuss] Analysis of patch 2e4ca00 for StarlingX Upstreaming

 

Hi Matt,

The patch 2e4ca00 changes function get_max_timeout() by changing the multiplication factor of 10 to 1 to avoid wait for 60*10 seconds(The default value for rpc_response_timeout is 60). This can be done by changing the default value from 60 to 6. Then will wait 6*10 seconds. This patch may be just a configuration problem to neutron upstream. Do you think we should upstream this patch or not?

 

Best Regards,

Xu, Chenjie