What I see in the data looks like slow clients. Note that the 99th and max
values for UA_READ_HEADER_DONE is almost as large as the same values for
UA_CLOSE. This would seem to mean the client finishes sending its header after
200s and then immediately closes the connection. I'd do a few packet traces on
the inbound side at the ATS box and see what is taking so long for the client
to send its request.
On Friday, January 22, 2016 6:53 PM, Adi Mallikarjuna Reddy V
<[email protected]> wrote:
Thanks for the setting info.
I have enabled it (with 30sec ) and recorded couple of such request. Here is
the summary as shown by the perl script. I looked at mean and found that the
time taken between ua_close to sm_finish is higher. Do we know what causes
this? Is this normal?
-bash-4.1$ ./slow_log_report.pl /home/y/logs/trafficserver/diags.log
key total count mean median 95th
99th min max ua_begin 0.00 46.00
0.00 0.00 0.00 0.00 0.00 0.00
ua_read_header_done 346.18 24.00 14.42 0.00 129.47
216.70 0.00 216.70 dns_lookup_begin 346.18 24.00
14.42 0.00 129.47 216.70 0.00 216.70
dns_lookup_end 346.19 24.00 14.42 0.00 129.47
216.70 0.00 216.70 server_connect 346.19 24.00
14.42 0.00 129.47 216.70 0.00 216.70
server_first_read 1333.69 23.00 57.99 45.11 130.21
216.94 30.12 216.94 server_read_header_done 1333.69 23.00
57.99 45.11 130.21 216.94 30.12 216.94
server_close 1261.34 21.00 60.06 45.36 130.21 216.94
30.12 216.94 ua_close 1634.74 24.00 68.11
45.36 216.94 300.99 30.12 300.99 sm_finish
6865.05 46.00 149.24 101.50 300.80 300.99 30.12
300.99
ThanksAdi
On Fri, Jan 22, 2016 at 1:46 PM, Scott Beardsley <[email protected]> wrote:
Look at this setting in records.config:
CONFIG proxy.config.http.slow.log.threshold INT N
see also
http://trafficserver.readthedocs.org/en/latest/admin-guide/files/records.config.en.html?highlight=proxy.config.http.slow.log.threshold#proxy-config-http-slow-log-threshold
This setting will log extra requests when they take more than N milliseconds.
You can use a perl script here to spit out some useful data:
https://github.com/apache/trafficserver/blob/master/tools/slow_log_report.pl
On Friday, January 22, 2016 1:41 PM, Adi Mallikarjuna Reddy V
<[email protected]> wrote:
I am newbie. Can you elaborate on enabling slow logging?
ThanksAdi
On Fri, Jan 22, 2016 at 7:58 AM, Alan Carroll <[email protected]>
wrote:
I might try enabling slow logging at around 20..25 seconds and see what part of
the transaction is slow enough to cause the client to abort. This should also
help determine if the client aborts are due to the clients or because the
transaction took too long.
On Wednesday, January 20, 2016 5:33 PM, Adi Mallikarjuna Reddy V
<[email protected]> wrote:
Hello,
We are seeing an higher number of ERR_CLIENT_ABORT/200 errors in our usage of
Traffic Server.
Our primary purpose of traffic server is only to act as Reverse Proxy and
mainly for SSL termination (without caching).
We are currently using:
Traffic Server version - 5.3.0 OS - Red Hat Enterprise Linux Server release 6.5
The following are the configuration parameters:
CONFIG proxy.config.http.keep_alive_no_activity_timeout_in INT 115CONFIG
proxy.config.http.keep_alive_no_activity_timeout_out INT 120CONFIG
proxy.config.http.transaction_no_activity_timeout_in INT 300CONFIG
proxy.config.http.transaction_no_activity_timeout_out INT 300CONFIG
proxy.config.http.transaction_active_timeout_in INT 900CONFIG
proxy.config.http.transaction_active_timeout_out INT 0CONFIG
proxy.config.http.accept_no_activity_timeout INT 300CONFIG
proxy.config.http.background_fill_active_timeout INT 120CONFIG
proxy.config.http.background_fill_completed_threshold FLOAT 0.500000CONFIG
proxy.config.http.connect_attempts_max_retries INT 0CONFIG
proxy.config.http.connect_attempts_max_retries_dead_server INT 3CONFIG
proxy.config.http.connect_attempts_rr_retries INT 3CONFIG
proxy.config.http.connect_attempts_timeout INT 320CONFIG
proxy.config.http.post_connect_attempts_timeout INT 1800CONFIG
proxy.config.http.down_server.cache_time INT 300CONFIG
proxy.config.http.down_server.abort_threshold INT 10
ThanksAdi
