Re: [apache/trafficserver] HTTP2 drain (#1710)

2017-06-13 Thread Masakazu Kitajo
On Tue, Jun 13, 2017 at 12:56 PM, James Peach wrote: > > > On 11 Jun 2017, at 4:45, Masakazu Kitajo wrote: > > On Sun, Jun 11, 2017 at 1:27 PM, James Peach wrote: >> >> >>> >>> On 9 May 2017, at 0:02, Masakazu Kitajo wrote: >>> >>> Hi, >>> Don't worry guys. I'm working on it.

Re: [apache/trafficserver] HTTP2 drain (#1710)

2017-06-12 Thread James Peach
On 11 Jun 2017, at 4:45, Masakazu Kitajo wrote: On Sun, Jun 11, 2017 at 1:27 PM, James Peach wrote: On 9 May 2017, at 0:02, Masakazu Kitajo wrote: Hi, Don't worry guys. I'm working on it. I merged the PR to have a draining (shedding) logic for HTTP/2 (Double GOAWAY frames), and now

Re: [apache/trafficserver] HTTP2 drain (#1710)

2017-06-11 Thread Masakazu Kitajo
On Sun, Jun 11, 2017 at 1:27 PM, James Peach wrote: > > > On 9 May 2017, at 0:02, Masakazu Kitajo wrote: > > Hi, >> >> Don't worry guys. I'm working on it. >> >> I merged the PR to have a draining (shedding) logic for HTTP/2 (Double >> GOAWAY frames), and now I'm generalizing the flag so that we

Re: [apache/trafficserver] HTTP2 drain (#1710)

2017-06-10 Thread James Peach
On 9 May 2017, at 0:02, Masakazu Kitajo wrote: Hi, Don't worry guys. I'm working on it. I merged the PR to have a draining (shedding) logic for HTTP/2 (Double GOAWAY frames), and now I'm generalizing the flag so that we can invoke draining on both HTTP/1 and HTTP/2. I'm going to add "Connec

Re: [apache/trafficserver] HTTP2 drain (#1710)

2017-05-09 Thread Masakazu Kitajo
On Tue, May 9, 2017 at 11:12 PM, James Peach wrote: > > > On 9 May 2017, at 0:02, Masakazu Kitajo wrote: > > Hi, >> >> Don't worry guys. I'm working on it. >> >> I merged the PR to have a draining (shedding) logic for HTTP/2 (Double >> GOAWAY frames), and now I'm generalizing the flag so that we

Re: [apache/trafficserver] HTTP2 drain (#1710)

2017-05-09 Thread James Peach
On 9 May 2017, at 0:02, Masakazu Kitajo wrote: Hi, Don't worry guys. I'm working on it. I merged the PR to have a draining (shedding) logic for HTTP/2 (Double GOAWAY frames), and now I'm generalizing the flag so that we can invoke draining on both HTTP/1 and HTTP/2. I'm going to add "Connec

Re: [apache/trafficserver] HTTP2 drain (#1710)

2017-05-09 Thread Masakazu Kitajo
Hi, Don't worry guys. I'm working on it. I merged the PR to have a draining (shedding) logic for HTTP/2 (Double GOAWAY frames), and now I'm generalizing the flag so that we can invoke draining on both HTTP/1 and HTTP/2. I'm going to add "Connection: close" header automatically for HTTP/1. So, "pr

Re: [apache/trafficserver] HTTP2 drain (#1710)

2017-05-08 Thread CrazyCow
I don't disagree. The reasons I chose this way are: 1. We are using other stuff in my team instead of traffic_ctl to manage the process and do the upgrade. 2. traffic_ctl --drain can only support HTTP and it can only be used when restarting ATS. That makes it hardly useful in our use case. 2017-05

Re: [apache/trafficserver] HTTP2 drain (#1710)

2017-05-08 Thread James Peach
On 8 May 2017, at 21:48, Miles Libbey wrote: We'd also like a bit more fine grained control in the process -- we frequently want to perform maintenance on a server (upgrading ATS; upgrading the OS; performing hardware changes, etc) after draining but before restarting ATS. I suppose this would

Re: [apache/trafficserver] HTTP2 drain (#1710)

2017-05-08 Thread Miles Libbey
We'd also like a bit more fine grained control in the process -- we frequently want to perform maintenance on a server (upgrading ATS; upgrading the OS; performing hardware changes, etc) after draining but before restarting ATS. I suppose this would mean allowing the --drain option to apply to traf

Re: [apache/trafficserver] HTTP2 drain (#1710)

2017-05-08 Thread James Peach
This patch adds another separate shutdown mechanism that only works for HTTP/2. I think that we really ought to have a single, well-defined graceful shutdown that works for all protocols. In HTTP/1.1 it works like this: - You (possibly dynamically) set proxy.config.restart.active_client_thres