[SAC] Trac slow to respond

[Alex Mandel]

On 11/28/2016 09:56 AM, Markus Neteler wrote:
> On Mon, Nov 28, 2016 at 4:13 PM, Sandro Santilli <strk at kbt.io> wrote:
>> On Mon, Nov 28, 2016 at 03:09:46PM +0100, Markus Neteler wrote:
>>> On Nov 28, 2016 2:55 PM, "Jeff McKenna" <jmckenna at gatewaygeomatics.com>
>>> wrote:
>>>>
>>>> Possibly only for me, but I thought I'd report that Trac is very slow
> to
>>> load at the moment: https://trac.osgeo.org/geotiff/
>>>
>>> This is a known issue... Probably the apache config needs a review (see
> the
>>> earlier discussion from some days ago).
>>>
>>> I checked the server-status page on the machine but don't know what to
> tune.
>>
>> Is the status page any useful at understanding *which* service is
>> being slow ?
>
> Probably (BTW: I made a screenshot and then used "gocr" to be able to
> copy-paste it here. IPs xx'ed and also version numbers):
>
> neteler at tracsvn:~$ links http://localhost/server-status
>
>
> Apache Server Status for localhost
>
>  Server Version: Apache/2.xxxx [version omitted for the mailing list
> archive]
>
>  Server Built: [version omitted for the mailing list archive]
>  Current Time: monday, 28-Nov-2016 09:49:03 PST
>  Restart Time: Saturday, 26-Nov-2016 06:09:06 PST
>  Parent Server Generation: 1
>  Server uptime: 2 days 3 hours 39 minutes 56 seconds
>  Total accesses: 798705 - Total Traffic: 45.1 GB
>  CPU Usage: u47.77 s3.01 cu.12 cs0 - .0274_ CPU load
>  4.29 requests/sec - 254.0 kB/second - 59.2 kB/request
>  5 requests currently being processed, 10 idle workers
>
> _K_...._WC_C_K_.............................................
>
>  Scoreboard Key:
>  ___ Waiting for Connection, _S_ Starting up, _n_ Reading Request,
>  _W_ Sending Reply, _K_ Keepalive ( read) , _D_ DNS Lookup,
>  _C_ Closing connection, _L_ Logging, _G_ Gracefully finishing,
>  _I_ Idle cleanup of worker, _. _ Open slot with no current process
>
>  Srv PID  Acc       m CPU SS Req Conn Child Slot   Client        VHost
>    Request
>  0-1  24347 0/11/42190 _ 0.02 0  165  0.0  0.00  1926.83 xx.74.170.ww
> git.osgeo.org  NULL
>  1-1  24364 1/1/41394 K 1.30 0  1325  0.2  0.00  1972.41 xx.143.71.ww
> trac.osgeo.org GET /openlayers/chrome/common/collapsed.png HTTP/1 1
>  2-1  24288 0/7/40752  _ 2.37 4  159  0.0  0.06  1633.89 xx.143.71.ww
> git.osgeo.org  NULL
>  3-1  -     0/0/41000    4.22 1  61   0.0  0.00  1090.36 xx.143.71.ww
> git.osgeo.org  NULL
>  9-1  -     0/0/39910    2.67 1  69   0.0  0.00  3059.53 xx.143.71.ww
> git.osgeo.org  NULL
>  5-1  -     0/0/39344    3.27 0  1972 0.0  0.00  2294.13 xx.46.52.zz
> trac.osgeo.org GET
> /geos/timeline?from=2013-06-27T06_3A22_3A01-07_3A00 at precisi
>  6-1  -     0/0/38860    2.46 1  63   0.0  0.00  1726.96 xx.143.71.ww
> git.osgeo.org  NULL
>  7-1  24293 0/4/39105  _ 1.33 1  2    0.0  0.01  1035.16 xx.255.250.pp
> git.osgeo.org  NULL
>  8-1  24350 0/1/38107 W  1.85 0  0    0.0  0.01  2328.70 127.0.0.1
> git.osgeo.org  GET /server-status HTTP/1.1
>  9-1  24351 1/8/37419 C  1.95 0  1998 9.8  0.01  1647.76 xx.125.71.hh
> trac.osgeo.org GET /gdal/log/trunk/gdal/port/cpl_vsil.cpp?rev=30823 HTTP/1 1
>  10-1 24353 0/2/35222    0.00 1  0    0.0  0.00  3315.85 xx.154.133.ee
> git.osgeo.org  NULL
>
>
> In this very moment the load level is low.
>
> I suspect that the git.osgeo.org traffic not not sufficiently supported
> with the current apache configuration.
>
>> One thing I know I did for sure was to (long ago) raising logging
>> levels. Maybe reducing them would help a bit ? See /etc/trac and
>> /var/log/trac
>
> Could also be!
>
> Markus

Are all the apache logs in one log? If so, download the last rotate and 
do some parsing and counting of what's being hit (simple graphs grouped 
by the trac instance top level name should work). If they are separate 
logs, then just see which one is significantly bigger.

Is the git on the same machine, should be easy enough to send that to 
it's own log if it's a separate vhost.

-Alex