[SSI] Server does not call Finished() on client disconnect

[slithy]

In the normal case, when a client makes a request, Finished() is called before the client disconnects:

171103 11:36:34 14217 XrootdXeq: mike.14362:29@localhost pvt IP64 login
[DEBUG] Service::Prepare():
[DEBUG]    Resource name: /ctafrontend
[DEBUG] RequestProc() constructor
[DEBUG] XrdSsiPbService::ProcessRequest(): Binding Processor to Request
[DEBUG] RequestProc::Execute()
[DEBUG] RequestProc::Finished()
[DEBUG] XrdSsiPbService::ProcessRequest(): Unbinding Processor from Request
[DEBUG] ~RequestProc() destructor
171103 11:36:35 14217 XrootdXeq: mike.14362:29@localhost disc 0:00:01

However, if the client disconnects (process is killed or the connection times out), Finished() is not called until the server has finished processing the response. Here I artifically made the response take 100 seconds to illustrate the problem:

171103 11:36:54 14075 XrootdXeq: mike.14377:27@localhost pvt IP64 login
[DEBUG] Service::Prepare():
[DEBUG]    Resource name: /ctafrontend
[DEBUG] RequestProc() constructor
[DEBUG] XrdSsiPbService::ProcessRequest(): Binding Processor to Request
[DEBUG] RequestProc::Execute()
171103 11:36:55 14217 XrootdXeq: mike.14377:27@localhost disc 0:00:01
171103 11:38:34 14075 XrootdsendResp: mike.14377:27@localhost query async resp aborted; user gone.
[DEBUG] RequestProc::Finished()
[DEBUG] XrdSsiPbService::ProcessRequest(): Unbinding Processor from Request
[DEBUG] ~RequestProc() destructor

Note that the client disconnects at 11:36:55 but the response is not aborted immediately. It runs to completion before it realises that the client is no longer present. At 11:38:34, it realises that the user is gone and then calls Finished().

In the intervening time between the client disconnect and Finished(), new client connections are refused with this error:

Unable to write /ctafrontend; address already in use

(Note that normally multiple simultaneous connections are fine, it's only after a client disconnect that new connections are refused).

[xrootd-dev]

Hi Michael, Definitely a bug. Thanks for reporting it. I should be able to fix this before the weekend ends. Andy

…

On Fri, 3 Nov 2017, Michael wrote: In the normal case, when a client makes a request, Finished() is called before the client disconnects: ``` 171103 11:36:34 14217 XrootdXeq: ***@***.*** pvt IP64 login [DEBUG] Service::Prepare(): [DEBUG] Resource name: /ctafrontend [DEBUG] RequestProc() constructor [DEBUG] XrdSsiPbService::ProcessRequest(): Binding Processor to Request [DEBUG] RequestProc::Execute() [DEBUG] RequestProc::Finished() [DEBUG] XrdSsiPbService::ProcessRequest(): Unbinding Processor from Request [DEBUG] ~RequestProc() destructor 171103 11:36:35 14217 XrootdXeq: ***@***.*** disc 0:00:01 ``` However, if the client disconnects (process is killed or the connection times out), Finished() is not called until the server has finished processing the response. Here I artifically made the response take 100 seconds to illustrate the problem: ``` 171103 11:36:54 14075 XrootdXeq: ***@***.*** pvt IP64 login [DEBUG] Service::Prepare(): [DEBUG] Resource name: /ctafrontend [DEBUG] RequestProc() constructor [DEBUG] XrdSsiPbService::ProcessRequest(): Binding Processor to Request [DEBUG] RequestProc::Execute() 171103 11:36:55 14217 XrootdXeq: ***@***.*** disc 0:00:01 171103 11:38:34 14075 XrootdsendResp: ***@***.*** query async resp aborted; user gone. [DEBUG] RequestProc::Finished() [DEBUG] XrdSsiPbService::ProcessRequest(): Unbinding Processor from Request [DEBUG] ~RequestProc() destructor ``` Note that the client disconnects at 11:36:55 but the response is not aborted immediately. It runs to completion before it realises that the client is no longer present. At 11:38:34, it realises that the user is gone and then calls Finished(). In the intervening time between the client disconnect and Finished(), new client connections are refused with this error: ``` Unable to write /ctafrontend; address already in use ``` (Note that normally multiple simultaneous connections are fine, it's only after a client disconnect that new connections are refused). -- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: #616 ######################################################################## Use REPLY-ALL to reply to list To unsubscribe from the XROOTD-DEV list, click the following link: https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-DEV&A=1