WRITE_COMPLETE event in http server open causes traffic server to abort #1930

vmamidi · 2017-05-17T20:23:38Z

Stack trace :
ink_abort(char const*, ...)
HttpSM::state_http_server_open(int, void*)
HttpSM::main_handler(int, void*)
write_to_net_io(NetHandler*, UnixNetVConnection*, EThread*)
NetHandler::mainNetEvent(int, Event*)
EThread::execute()

and corresponding error messages:
traffic_server[33593]: {0x2aaab6d0c700} ERROR: [HttpSM::state_http_server_open] Unknown event: 103
traffic_server[33593]: FATAL: HttpSM.cc:1798: failed assertion 0

We are seeing this in 7.0.0

The text was updated successfully, but these errors were encountered:

scw00 · 2017-05-17T22:30:30Z

Print history in HttpSM

vmamidi · 2017-05-18T00:12:32Z

i dont have any more information

scw00 · 2017-05-18T01:35:51Z

:(.....sad!!

bryancall · 2017-05-23T21:47:08Z

@vmamidi What version is this for?

vmamidi · 2017-05-23T22:12:06Z

we are seeing this on 7.0.0

bryancall · 2017-05-24T16:42:53Z

@vmamidi Can you post a better backtrace with line numbers in the code? bt full would be great!

vmamidi · 2017-06-01T00:31:54Z

@scw00 I was able to get the history. Here is the history for ATS on 7.1.x branch.
[0] =
fileline = 0x55a3621f7912
*0x55a3621f7912 = string(0x55a3621f7912, [HttpSM.cc:630])
event = 100
reentrancy = 2
[1] =
fileline = 0x55a3621f7d6d
*0x55a3621f7d6d = string(0x55a3621f7d6d, [HttpSM.cc:1339])
event = 60000
reentrancy = 3
[2] =
fileline = 0x55a3621f7d98
*0x55a3621f7d98 = string(0x55a3621f7d98, [HttpSM.cc:1379])
event = 60000
reentrancy = 3
[3] =
fileline = 0x55a3621f7d6d
*0x55a3621f7d6d = string(0x55a3621f7d6d, [HttpSM.cc:1339])
event = 60000
reentrancy = 4
[4] =
fileline = 0x55a3621f7d98
*0x55a3621f7d98 = string(0x55a3621f7d98, [HttpSM.cc:1379])
event = 60000
reentrancy = 4
[5] =
fileline = 0x55a3621f7d6d
*0x55a3621f7d6d = string(0x55a3621f7d6d, [HttpSM.cc:1339])
event = 60000
reentrancy = 5
[6] =
fileline = 0x55a3621f7d98
*0x55a3621f7d98 = string(0x55a3621f7d98, [HttpSM.cc:1379])
event = 60000
reentrancy = 5
[7] =
fileline = 0x55a3621f7d6d
*0x55a3621f7d6d = string(0x55a3621f7d6d, [HttpSM.cc:1339])
event = 60000
reentrancy = 6
[8] =
fileline = 0x55a3621f7d98
*0x55a3621f7d98 = string(0x55a3621f7d98, [HttpSM.cc:1379])
event = 60000
reentrancy = 6
[9] =
fileline = 0x55a3621f7d6d
*0x55a3621f7d6d = string(0x55a3621f7d6d, [HttpSM.cc:1339])
event = 60000
reentrancy = 7
[10] =
fileline = 0x55a3621f7d98
*0x55a3621f7d98 = string(0x55a3621f7d98, [HttpSM.cc:1379])
event = 60000
reentrancy = 7
[11] =
fileline = 0x55a3621fa4bc
*0x55a3621fa4bc = string(0x55a3621fa4bc, [HttpSM.cc:7373])
event = 65535
reentrancy = 7
[12] =
fileline = 0x55a3621f2ebd
*0x55a3621f2ebd = string(0x55a3621f2ebd, [HttpCacheSM.cc:118])
event = 1103
reentrancy = -1
[13] =
fileline = 0x55a3621f84d6
*0x55a3621f84d6 = string(0x55a3621f84d6, [HttpSM.cc:2577])
event = 1103
reentrancy = 8
[14] =
fileline = 0x55a3621f7d6d
*0x55a3621f7d6d = string(0x55a3621f7d6d, [HttpSM.cc:1339])
event = 60000
reentrancy = 9
[15] =
fileline = 0x55a3621f7d98
*0x55a3621f7d98 = string(0x55a3621f7d98, [HttpSM.cc:1379])
event = 60000
reentrancy = 9
[16] =
fileline = 0x55a3621f7d6d
*0x55a3621f7d6d = string(0x55a3621f7d6d, [HttpSM.cc:1339])
event = 60000
reentrancy = 10
[17] =
fileline = 0x55a3621f7d98
*0x55a3621f7d98 = string(0x55a3621f7d98, [HttpSM.cc:1379])
event = 60000
reentrancy = 10
[18] =
fileline = 0x55a3621fa49e
*0x55a3621fa49e = string(0x55a3621fa49e, [HttpSM.cc:7359])
event = 65535
reentrancy = 10
[19] =
fileline = 0x55a3621fa4e9
*0x55a3621fa4e9 = string(0x55a3621fa4e9, [HttpSM.cc:7488])
event = 65535
reentrancy = 10
[20] =
fileline = 0x55a3621f2f73
*0x55a3621f2f73 = string(0x55a3621f2f73, [HttpCacheSM.cc:177])
event = 1108
reentrancy = -1
[21] =
fileline = 0x55a3621f8429
*0x55a3621f8429 = string(0x55a3621f8429, [HttpSM.cc:2458])
event = 1108
reentrancy = 11
[22] =
fileline = 0x55a3621fa4da
*0x55a3621fa4da = string(0x55a3621fa4da, [HttpSM.cc:7386])
event = 65535
reentrancy = 11
[23] =
fileline = 0x55a3621f9263
*0x55a3621f9263 = string(0x55a3621f9263, [HttpSM.cc:4578])
event = 0
reentrancy = 7
[24] =
fileline = 0x55a3621f8055
*0x55a3621f8055 = string(0x55a3621f8055, [HttpSM.cc:1980])
event = 103
reentrancy = 1
[25] =
fileline = 0x55a3621f7fd2
*0x55a3621f7fd2 = string(0x55a3621f7fd2, [HttpSM.cc:1810])
event = 104
reentrancy = 1
[26] =
fileline = 0x55a3621f96cc
*0x55a3621f96cc = string(0x55a3621f96cc, [HttpSM.cc:5485])
event = 104
reentrancy = 1
[27] =
fileline = 0x55a3621fa4da
*0x55a3621fa4da = string(0x55a3621fa4da, [HttpSM.cc:7386])
event = 65535
reentrancy = 1
[28] =
fileline = 0x55a3621f8055
*0x55a3621f8055 = string(0x55a3621f8055, [HttpSM.cc:1980])
event = 103
reentrancy = 1
[29] =
fileline = 0x55a3621f7fd2
*0x55a3621f7fd2 = string(0x55a3621f7fd2, [HttpSM.cc:1810])
event = 104
reentrancy = 1
[30] =
fileline = 0x55a3621f96cc
*0x55a3621f96cc = string(0x55a3621f96cc, [HttpSM.cc:5485])
event = 104
reentrancy = 1
[31] =
fileline = 0x55a3621fa49e
*0x55a3621fa49e = string(0x55a3621fa49e, [HttpSM.cc:7359])
event = 65535
reentrancy = 1
[32] =
fileline = 0x55a3621f7fd2
*0x55a3621f7fd2 = string(0x55a3621f7fd2, [HttpSM.cc:1810])
event = 3
reentrancy = 1
[33] =
fileline = 0x55a3621f96cc
*0x55a3621f96cc = string(0x55a3621f96cc, [HttpSM.cc:5485])
event = 3
reentrancy = 1
[34] =
fileline = 0x55a3621fa4da
*0x55a3621fa4da = string(0x55a3621fa4da, [HttpSM.cc:7386])
event = 65535
reentrancy = 1
[35] =
fileline = 0x55a3621f7ef5
*0x55a3621f7ef5 = string(0x55a3621f7ef5, [HttpSM.cc:1716])
event = 200
reentrancy = 2
[36] =
fileline = 0x55a3621f8055
*0x55a3621f8055 = string(0x55a3621f8055, [HttpSM.cc:1980])
event = 103
reentrancy = 1
[37] =
fileline = 0x55a3621f7ef5
*0x55a3621f7ef5 = string(0x55a3621f7ef5, [HttpSM.cc:1716])
event = 500
reentrancy = 1

scw00 · 2017-06-01T00:35:03Z

great!

scw00 · 2017-06-01T01:18:26Z

Ok I got it.

The reason is that the HttpSM::state_send_server_request_header may receive the ERROR event when read side down. Then ats retry another server without cleaning the write side.

It may reference to #1629 and #1559 .

scw00 · 2017-06-01T01:22:28Z

@oknet

vmamidi · 2017-06-01T02:24:20Z

looks like HttpSM is receiving VC_EVENT_ERROR and also WRITE_COMPLETE event after EOS.

oknet · 2017-06-01T04:50:46Z

ATS do a hostdb lookup to retry the next ip by round robin.

[31] =
fileline = 0x55a3621fa49e
*0x55a3621fa49e = string(0x55a3621fa49e, [HttpSM.cc:7359])
event = 65535
reentrancy = 1

The hostdb callback at [37]

[37] =
fileline = 0x55a3621f7ef5
*0x55a3621f7ef5 = string(0x55a3621f7ef5, [HttpSM.cc:1716])
event = 500
reentrancy = 1

I don't know the reason why [32] shows in the history, it should blocked by pending_action.

Also, I noticed :

traffic_server[33593]: {0x2aaab6d0c700} ERROR: [HttpSM::state_http_server_open] Unknown event: 103

The event is changed from 500 to 103. What is that means ?
May be one callback to HttpSM without get the mutex lock first ? HostDB bug ?

oknet · 2017-06-01T04:54:59Z

@scw00 I think we should cleanup server_entry at HttpSM.cc:7359, just do the same thing at HttpSM.cc:7386.

oknet · 2017-06-01T05:02:06Z

@vmamidi Did you seen "Unknown event: 500" ?

vmamidi · 2017-06-01T05:17:33Z

@oknet yes, it is different from the original stack trace that I originally posted but it still falls under the same category of receiving the wrong event.
Log is
traffic_server[3732]: {0x2b23e4f0c700} ERROR: [HttpSM::state_http_server_open] Unknown event: 500

and the Stack trace is:

abort
ink_abort
HttpSM::state_http_server_open(int, void*)
HttpSM::main_handler(int, void*)
handleEvent
reply_to_cont
HostDBContinuation::dnsEvent(int, HostEnt*)
DNSEntry::postEvent(int, Event*)
EThread::process_event(Event*, int)
EThread::execute()
spawn_thread_internal

oknet · 2017-06-01T05:31:04Z

@vmamidi I think I found the reason, a patch will be pushed later. Can you reproduce it and have a try ?

vmamidi · 2017-06-01T05:41:32Z

what do you think is the problem?

oknet · 2017-06-01T05:57:00Z

@vmamidi The HttpSM retry to connect to other server IPs by DNS rr. (The [31] shows)
The HttpSM should waiting for HostDB callback and go on.
But the server_vc is still not closed, then the NetHandler got an error from the socket fd and callback to HttpSM. (The [32] shows)

From the HttpSM::history, the [32] changes the state of HttpSM and reset the default handler of HttpSM.
It makes the HttpSM goes to wrong state and the HostDB callback goes to wrong default handler.

vmamidi · 2017-06-01T06:02:14Z

@oknet HttpSM::handle_server_setup_error at [30] should clear all of that

oknet · 2017-06-01T06:03:15Z

@vmamidi No, the [30] does not cleanup server entry.

scw00 · 2017-06-01T06:04:51Z

Can we do clean in HttpSM::handle_server_setup_error ?

oknet · 2017-06-01T06:11:17Z

@scw00 I don't know. :-(

vmamidi · 2017-06-01T06:20:36Z

@oknet @scw00
I don't think this is related to just DNS.

look at this stack trace.

abort
ink_abort(char const*, ...)
HttpSM::state_http_server_open(int, void*)
HttpSM::main_handler(int, void*)
write_to_net_io(NetHandler*, UnixNetVConnection*, EThread*)
NetHandler::mainNetEvent(int, Event*)
EThread::execute()

and corresponding log message.
traffic_server[29074]: {0x2aaab730c700} ERROR: [HttpSM::state_http_server_open] Unknown event: 103

vmamidi · 2017-06-01T06:21:38Z

Also, i am not pretty sure but i dont think this was an issue in earlier versions of ATS.

oknet · 2017-06-01T06:23:02Z

@vmamidi
In the [31], the default handler is set to HttpSM::state_hostdb_lookup.

The NetHandler callback ERROR to HttpSM [32] and then run into [33] and [34].
In the [34], the default handler changed by:

7250   case HttpTransact::SM_ACTION_ORIGIN_SERVER_OPEN: {
7251     if (congestionControlEnabled && (t_state.congest_saved_next_action == HttpTransact::SM_ACTION_UNDEFINED)) {
7252       t_state.congest_saved_next_action = HttpTransact::SM_ACTION_ORIGIN_SERVER_OPEN;
7253       HTTP_SM_SET_DEFAULT_HANDLER(&HttpSM::state_congestion_control_lookup);
7254       if (!do_congestion_control_lookup()) {
7255         break;
7256       }
7257     }
7258     HTTP_SM_SET_DEFAULT_HANDLER(&HttpSM::state_http_server_open);

In the [37], the HostDB callback to HttpSM and the default handler should be HttpSM::state_hostdb_lookup but it is changed to HttpSM::state_http_server_open.

oknet · 2017-06-01T06:25:41Z

@vmamidi Yes, only after #947 merged.

scw00 · 2017-06-01T06:33:23Z

Hmm! In my opinion, we should clean up the server entry when we decide to try again, regardless of whether the server crashes.

vmamidi · 2017-06-01T06:33:45Z

@oknet yes, i agree with related to history that i have pasted here but i dont think that is the only one of the scenarios.

vmamidi · 2017-06-01T06:33:57Z

@scw00 yes, i agreee

vmamidi · 2017-06-01T07:05:32Z

@oknet here is another history.

        [0] = 
           fileline = 0x2aaaaaf2caef
              *0x2aaaaaf2caef = string(0x2aaaaaf2caef, [HttpSM.cc:628])
           event = 100
           reentrancy = 2
        [1] = 
           fileline = 0x2aaaaaf2d147
              *0x2aaaaaf2d147 = string(0x2aaaaaf2d147, [HttpSM.cc:1337])
           event = 60000
           reentrancy = 3
        [2] = 
           fileline = 0x2aaaaaf2ce6a
              *0x2aaaaaf2ce6a = string(0x2aaaaaf2ce6a, [HttpSM.cc:1377])
           event = 60000
           reentrancy = 3
        [3] = 
           fileline = 0x2aaaaaf2d147
              *0x2aaaaaf2d147 = string(0x2aaaaaf2d147, [HttpSM.cc:1337])
           event = 60000
           reentrancy = 4
        [4] = 
           fileline = 0x2aaaaaf2ce6a
              *0x2aaaaaf2ce6a = string(0x2aaaaaf2ce6a, [HttpSM.cc:1377])
           event = 60000
           reentrancy = 4
        [5] = 
           fileline = 0x2aaaaaf2d147
              *0x2aaaaaf2d147 = string(0x2aaaaaf2d147, [HttpSM.cc:1337])
           event = 60000
           reentrancy = 5
        [6] = 
           fileline = 0x2aaaaaf2ce6a
              *0x2aaaaaf2ce6a = string(0x2aaaaaf2ce6a, [HttpSM.cc:1377])
           event = 60000
           reentrancy = 5
        [7] = 
           fileline = 0x2aaaaaf2d147
              *0x2aaaaaf2d147 = string(0x2aaaaaf2d147, [HttpSM.cc:1337])
           event = 60000
           reentrancy = 6
        [8] = 
           fileline = 0x2aaaaaf2ce6a
              *0x2aaaaaf2ce6a = string(0x2aaaaaf2ce6a, [HttpSM.cc:1377])
           event = 60000
           reentrancy = 6
        [9] = 
           fileline = 0x2aaaaaf2d147
              *0x2aaaaaf2d147 = string(0x2aaaaaf2d147, [HttpSM.cc:1337])
           event = 60000
           reentrancy = 7
        [10] = 
           fileline = 0x2aaaaaf2ce6a
              *0x2aaaaaf2ce6a = string(0x2aaaaaf2ce6a, [HttpSM.cc:1377])
           event = 60000
           reentrancy = 7
        [11] = 
           fileline = 0x2aaaaaf2d046
              *0x2aaaaaf2d046 = string(0x2aaaaaf2d046, [HttpSM.cc:7260])
           event = 65535
           reentrancy = 7
        [12] = 
           fileline = 0x2aaaaaf257e1
              *0x2aaaaaf257e1 = string(0x2aaaaaf257e1, [HttpCacheSM.cc:118])
           event = 1103
           reentrancy = -1
        [13] = 
           fileline = 0x2aaaaaf2cfdf
              *0x2aaaaaf2cfdf = string(0x2aaaaaf2cfdf, [HttpSM.cc:2573])
           event = 1103
           reentrancy = 8
        [14] = 
           fileline = 0x2aaaaaf2d147
              *0x2aaaaaf2d147 = string(0x2aaaaaf2d147, [HttpSM.cc:1337])
           event = 60000
           reentrancy = 9
        [15] = 
           fileline = 0x2aaaaaf2ce6a
              *0x2aaaaaf2ce6a = string(0x2aaaaaf2ce6a, [HttpSM.cc:1377])
           event = 60000
           reentrancy = 9
        [16] = 
           fileline = 0x2aaaaaf2d147
              *0x2aaaaaf2d147 = string(0x2aaaaaf2d147, [HttpSM.cc:1337])
           event = 60000
           reentrancy = 10
        [17] = 
           fileline = 0x2aaaaaf2ce6a
              *0x2aaaaaf2ce6a = string(0x2aaaaaf2ce6a, [HttpSM.cc:1377])
           event = 60000
           reentrancy = 10
        [18] = 
           fileline = 0x2aaaaaf2d028
              *0x2aaaaaf2d028 = string(0x2aaaaaf2d028, [HttpSM.cc:7246])
           event = 65535
           reentrancy = 10
        [19] = 
           fileline = 0x2aaaaaf2d028
              *0x2aaaaaf2d028 = string(0x2aaaaaf2d028, [HttpSM.cc:7246])
           event = 65535
           reentrancy = 10
        [20] = 
           fileline = 0x2aaaaaf2d073
              *0x2aaaaaf2d073 = string(0x2aaaaaf2d073, [HttpSM.cc:7375])
           event = 65535
           reentrancy = 10
        [21] = 
           fileline = 0x2aaaaaf2581d
              *0x2aaaaaf2581d = string(0x2aaaaaf2581d, [HttpCacheSM.cc:177])
           event = 1108
           reentrancy = -1
        [22] = 
           fileline = 0x2aaaaaf2cac3
              *0x2aaaaaf2cac3 = string(0x2aaaaaf2cac3, [HttpSM.cc:2454])
           event = 1108
           reentrancy = 11
        [23] = 
           fileline = 0x2aaaaaf2d064
              *0x2aaaaaf2d064 = string(0x2aaaaaf2d064, [HttpSM.cc:7273])
           event = 65535
           reentrancy = 11
        [24] = 
           fileline = 0x2aaaaaf2c6e4
              *0x2aaaaaf2c6e4 = string(0x2aaaaaf2c6e4, [HttpSM.cc:4545])
           event = 0
           reentrancy = 7
        [25] = 
           fileline = 0x2aaaaaf2ceb2
              *0x2aaaaaf2ceb2 = string(0x2aaaaaf2ceb2, [HttpSM.cc:1978])
           event = 103
           reentrancy = 1
        [26] = 
           fileline = 0x2aaaaaf2c7c6
              *0x2aaaaaf2c7c6 = string(0x2aaaaaf2c7c6, [HttpSM.cc:5621])
           event = 65535
           reentrancy = 1
        [27] = 
           fileline = 0x2aaaaaf2c90c
              *0x2aaaaaf2c90c = string(0x2aaaaaf2c90c, [HttpSM.cc:3465])
           event = 105
           reentrancy = 0
        [28] = 
           fileline = 0x2aaaaaf2ceb2
              *0x2aaaaaf2ceb2 = string(0x2aaaaaf2ceb2, [HttpSM.cc:1978])
           event = 105
           reentrancy = 1
        [29] = 
           fileline = 0x2aaaaaf2cae0
              *0x2aaaaaf2cae0 = string(0x2aaaaaf2cae0, [HttpSM.cc:5384])
           event = 105
           reentrancy = 1
        [30] = 
           fileline = 0x2aaaaaf2c588
              *0x2aaaaaf2c588 = string(0x2aaaaaf2c588, [HttpSM.cc:3585])
           event = 105
           reentrancy = 1
        [31] = 
           fileline = 0x2aaaaaf2cf0b
              *0x2aaaaaf2cf0b = string(0x2aaaaaf2cf0b, [HttpSM.cc:2743])
           event = 2301
           reentrancy = 2
        [32] = 
           fileline = 0x2aaaaaf2ced0
              *0x2aaaaaf2ced0 = string(0x2aaaaaf2ced0, [HttpSM.cc:5272])
           event = 0
           reentrancy = 2
        [33] = 
           fileline = 0x2aaaaaf2d064
              *0x2aaaaaf2d064 = string(0x2aaaaaf2d064, [HttpSM.cc:7273])
           event = 65535
           reentrancy = 2
        [34] = 
           fileline = 0x2aaaaaf2cf8c
              *0x2aaaaaf2cf8c = string(0x2aaaaaf2cf8c, [HttpSM.cc:1714])
           event = 103

Stack trace:

abort
ink_abort
HttpSM::state_http_server_open(int, void*)
HttpSM::main_handler(int, void*)
handleEvent
write_signal_and_update
write_signal_done(int, NetHandler*, UnixNetVConnection*)
write_to_net_io
NetHandler::mainNetEvent(int, Event*)

and corresponding log message:

traffic_server[33674]: {0x2aaab470c700} ERROR: [HttpSM::state_http_server_open] Unknown event: 103

oknet · 2017-06-01T09:06:57Z

@vmamidi What version is the last HttpSM::history for?

oknet · 2017-06-01T09:15:10Z

[25] WRITE_COMPLETE
[27] TIMEOUT
[31] HTTP_TUNNEL_EVENT_DONE

Is this a POST request ?

This is a retry after transform finished if it is a POST request with post transform enabled.
It is a different issue.

The PR #2034 is only for the first HttpSM::history.

scw00 · 2017-06-01T09:18:21Z

The post retry has broken currently because tunnel has consumed all post data when os down. So we can see [27] TIMEOUT.

oknet · 2017-06-01T09:27:38Z

The transformed data is dropped by HttpSM if server failed. The HttpSM can not rebuild the tunnel chain if retry to connect server.

The server connection retry mechanism has some bug, my suggestion is set the retry to 0 to disable the mechanism.

proxy.config.http.connect_attempts_max_retries INT 0
proxy.config.http.connect_attempts_max_retries_dead_server INT 0

If congestion.config is used then:

live_os_conn_retries and dead_os_conn_retries INT 0

vmamidi · 2017-06-01T13:45:32Z

@oknet it is 7.0.0.

Yes, ATS connection retries are broken and could cause a crash. This is the reason why I was saying having a fix while doing DNS which only solve one scenario.

scw00 · 2017-06-05T05:49:06Z

In the second history, we can see that client post timeout ([27]) and we send the 408 to client. Then server side will receive timeout and retry [28].

The problem is we did not close the vc directly.

That was fixed in 8.0.x

oknet · 2017-06-05T06:49:37Z

@vmamidi please try to fix the 2nd scenario by backport #1583 .

vmamidi · 2017-06-05T18:21:35Z

@oknet @scw00 @zwoop I am okay with having patches for each of the histories that we have on this issue but I am worried that state machine can receive events for past states that too because of something we did in #947 as pointed out by @oknet. As of now, I don't think state machine is robust enough to handle events for past states but if that is expected behavior we should make sure that HttpSM is robust enough to handle that.

vmamidi · 2017-06-05T22:00:58Z

@oknet @scw00 @zwoop @bryancall @SolidWallOfCode I understand the scenario mentioned in #947 but, state machine now can receive unexpected events. Now, we have to either fix the state machine with the new expectation or revert #947 and come up with a different solution which does not break the expectation and fixes the mentioned scenario. I personally do not think the current approach of leaving the state machine in not so robust state is a good idea.

zwoop · 2017-06-05T23:08:08Z

This might be worthwhile to take to the mailing list (dev@trafficserver). I'm ok either way, but I think it'd be better discussed / decided on the mailing list (I'm not sure who, if any, actually read these Issue comments :-).

zwoop · 2017-06-07T22:19:42Z

Everyone: I think we're to the point now of considering backing out #947. Is this agreeable by everyone?

scw00 · 2017-06-08T02:00:59Z

Agree!

oknet · 2017-06-08T14:35:34Z

+1

This reverts PRs apache#1559, apache#1522 and apache#947 PR apache#947 made the HTTP state machine unstable and lead to crashes in production like apache#1930 apache#1559 apache#1522 apache#1531 apache#1629 This reverts commit c1ac5f8.

This reverts PRs #1559, #1522 and #947 PR #947 made the HTTP state machine unstable and lead to crashes in production like #1930 #1559 #1522 #1531 #1629 This reverts commit c1ac5f8.

zwoop · 2017-06-23T20:31:16Z

I'm removing the 7.1.0 Milestone from this Issue, but keeping this open for now as there's valuable discussions here.

bryancall · 2019-05-22T17:50:30Z

If this is still an issue please reopen.

bryancall added the HTTP label May 23, 2017

bryancall added the Crash label May 24, 2017

oknet mentioned this issue Jun 1, 2017

Cleanup server_entry before do dns lookup #2034

Merged

zwoop added this to the 7.1.0 milestone Jun 6, 2017

zwoop added this to Critical issues in 7.x releases Jun 6, 2017

vmamidi mentioned this issue Jun 8, 2017

Revert PR #947 #2093

Merged

zwoop removed this from Critical issues in 7.x releases Jun 8, 2017

vmamidi mentioned this issue Jun 8, 2017

Revert PR #947 - merge to 7.1.x #2103

Merged

zwoop pushed a commit that referenced this issue Jun 8, 2017

Revert PR #947 - merge to 7.1.x

65a001b

This reverts PRs #1559, #1522 and #947 PR #947 made the HTTP state machine unstable and lead to crashes in production like #1930 #1559 #1522 #1531 #1629 This reverts commit c1ac5f8.

zwoop removed this from the 7.1.0 milestone Jun 23, 2017

bryancall closed this as completed May 22, 2019

WRITE_COMPLETE event in http server open causes traffic server to abort #1930

WRITE_COMPLETE event in http server open causes traffic server to abort #1930

Comments

vmamidi commented May 17, 2017 • edited

scw00 commented May 17, 2017

vmamidi commented May 18, 2017

scw00 commented May 18, 2017

bryancall commented May 23, 2017

vmamidi commented May 23, 2017

bryancall commented May 24, 2017

vmamidi commented Jun 1, 2017

scw00 commented Jun 1, 2017

scw00 commented Jun 1, 2017 • edited

scw00 commented Jun 1, 2017

vmamidi commented Jun 1, 2017

oknet commented Jun 1, 2017

oknet commented Jun 1, 2017

oknet commented Jun 1, 2017

vmamidi commented Jun 1, 2017

oknet commented Jun 1, 2017

vmamidi commented Jun 1, 2017

oknet commented Jun 1, 2017

vmamidi commented Jun 1, 2017

oknet commented Jun 1, 2017

scw00 commented Jun 1, 2017

oknet commented Jun 1, 2017

vmamidi commented Jun 1, 2017

vmamidi commented Jun 1, 2017

oknet commented Jun 1, 2017

oknet commented Jun 1, 2017

scw00 commented Jun 1, 2017

vmamidi commented Jun 1, 2017

vmamidi commented Jun 1, 2017

vmamidi commented Jun 1, 2017

oknet commented Jun 1, 2017

oknet commented Jun 1, 2017

scw00 commented Jun 1, 2017

oknet commented Jun 1, 2017

vmamidi commented Jun 1, 2017

scw00 commented Jun 5, 2017 • edited

oknet commented Jun 5, 2017

vmamidi commented Jun 5, 2017

vmamidi commented Jun 5, 2017

zwoop commented Jun 5, 2017

zwoop commented Jun 7, 2017

scw00 commented Jun 8, 2017

oknet commented Jun 8, 2017

zwoop commented Jun 23, 2017

bryancall commented May 22, 2019

vmamidi commented May 17, 2017 •

edited

scw00 commented Jun 1, 2017 •

edited

scw00 commented Jun 5, 2017 •

edited