 |
 |
On 08/31/13 21:25, Andrew Hanushevsky wrote: > Hi Matvz, > > On Sat, 31 Aug 2013, Matevz Tadel wrote: >> Thanks for bearing with us through all this thread :) > Actually, thank you for bearing with my vacation :-) > >> Keep in mind that if we force creation of sockets for each client, it might not >> make much sense to keep them alive ... or old, non-used sockets can get reused >> by new clients? Will this be properly reported in the monitoring 'u' stream? > I do agree, for long-lived servers we do need an idle time-out on the client. We > can't depend that servers will have their timeouts enabled and if they do that > the timeout value is decent for the proxy server. The non-used sockets cannot be > easily resused because they are tied to a security context so you would need to > figure out if there are any conflicts. So, I'd say that they can't be > practically reused. If they are not resused then the 'u' stream question is > immaterial. Yes, I somewhat expected that ... why can't the client just close them when it is destroyed then :) >> Andy, within the discussion I also asked for 't' stream to be flushed on close >> event. This way monitoring can be done with the file even though the session >> still remains opened for unknown period of time. > That's a pretty hefty request. Let me look into it. This shouldn't be super hard ... you are already flushing the buffers if time difference from initial frame is greater than some delta_t ... so the machinery should all be there and just needs to be put in the "close handler". Remember, this is only when io (or iov) is in effect and each session gets independent monitoring buffer/stream. Matevz ######################################################################## 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