754689 - Need access to personas and lightweight themes log

Status: RESOLVED FIXED

(Cloud Services :: Operations: Marketplace, task)

Description

[Pedro Alves]

We process the lightweight themes and personas web logs to count the number of users. However, after the recent move to scl3, those etl jobs have been failing / undercounting by a large margin. We need to find where the new logs are.


Example of a personas log entry we used to parse from /mnt/stats_im-log03/stats/logs/im-log01/www.getpersonas.com/ :


174.35.64.50 www.getpersonas.com - [15/Apr/2012:00:00:05 -0700] "GET /static/index_1.json?type=current&id=33&duration=228&appID=%7Bec8030f7-c20a-464f-9b0e-13a3a9e97384%7D&appVersion=11.0&appLocale=sv-SE&appOS=WINNT&appABI=x86-msvc HTTP/1.1" 200 114112 "-" "Mozilla/5.0 (Windows NT 6.1; rv:11.0) Gecko/20100101 Firefox/11.0" "-"


Example from a lightweight themes entry we used to parse from /mnt/stats_im-log03/stats/logs/im-log01/www.getpersonas.com-ssl :

217.164.100.144 www.getpersonas.com - [09/May/2012:11:01:24 -0700] "GET /en-US/update_check/200584 HTTP/1.1" 500 630 "-" "Mozilla/5.0 (Windows NT 6.0; rv:12.0) Gecko/20100101 Firefox/12.0" "-"



We've been told that we should look in /mnt/stats_im-log02/stats/logs/zlb5.ops.scl3.mozilla.com/ , and there are indeed a bunch of files there re personas. But... only after we spent a bunch of cpu processing time here we noticed there's not a single entry like the above.


Where can we find them?

Comment 1

[Jason Thomas [:jason]]

There was an error in log format in Zeus. All the logs since the move to SCL3 are missing the HTTP request URL due undefined variable in the log format. Because this variable was undefined, zeus was replacing the URL with '-'.

This should be fixed now and I am seeing requests that are similar to the ones mentioned in comment 0. The logs should be accessible at /mnt/stats_im-log02/stats/logs/zlb5.ops.scl3.mozilla.com/www.getpersonas.com.access_*

www.getpersonas.com.access_2012-05-14-07* should be the first set of logs with the corrected log format. Please verify and let me know if you continue to have issues.

Comment 2

[mroldan]

Jason,
The lightweight themes entries look fine now.
But we still miss the entries for personas. There are very few of them.
For ex. in April we had an average of 30 requests /sec.
In the logs for today, we have less than 5 request /hour.

Comment 3

[Jason Thomas [:jason]]

(In reply to Maria Roldan from comment #2)
> Jason,
> The lightweight themes entries look fine now.
> But we still miss the entries for personas. There are very few of them.
> For ex. in April we had an average of 30 requests /sec.
> In the logs for today, we have less than 5 request /hour.

I am not certain how we are parsing personas logs but doing a quick grep for 
'index_1.json?' in today's personas logs shows the following amount of requests/ hour:
for i in  www.getpersonas.com.access_2012-05-15-0?.gz; do echo $i; zgrep 'index_1.json?' $i | wc -l ; done
www.getpersonas.com.access_2012-05-15-00.gz
76553
www.getpersonas.com.access_2012-05-15-01.gz
68851
www.getpersonas.com.access_2012-05-15-02.gz
64869
www.getpersonas.com.access_2012-05-15-03.gz
69126
www.getpersonas.com.access_2012-05-15-04.gz
82255

If there is some other way I can test or look please let me know. Please also feel free to ping me on irc :jason.

Comment 4

[mroldan]

Thanks, Jason.
I think I know what the problem is. Since the logs changed, we started to filter lines with 'http://' por personas and lines with 'https://' for lightweight_themes. Then, we have this:

(1) SAMPLE LINE WITH 'https://' (lines we keep in lightweight_themes after initial filter)
68.148.48.54 www.getpersonas.com - [14/May/2012:17:12:36 -0700] 
"GET /en-US/update_check/45001 HTTP/1.1" 200 1589 "https://www.getpersonas.com/" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10.5; rv:11.0) Gecko/20100101 Firefox/11.0" ....etc.

(2) SAMPLE LINE WITH 'http://' (lines we keep in personas after initial filter)
14.0.33.198 www.getpersonas.com - [14/May/2012:17:00:19 -0700] 
"GET /static_json/6/4/44964/preview.jpg?1260925626 HTTP/1.1" 304 210 "http://www.getpersonas.com/ja/gallery/Film%20and%20TV/Movers/2" ...etc.

(3) SAMPLE LINE WITHOUT FILTER 
68.232.36.104 www.getpersonas.com - [14/May/2012:17:00:19 -0700] 
"GET /static/index_1.json?type=current&id=33&duration=1643&appID=%7Bec8030f7-c20a-464f-9b0e-13a3a9e97384%7D&appVersion=12.0&appLocale=en-US&appOS=WINNT&appABI=x86-msvc HTTP/1.1" 200 324376 "-" ...etc.


This way, the lines we keep for personas (2) don't match our regular expression. But the lines like (3) do. 

I think that, for personas, we shouldn't keep the lines containing 'http://' but the lines that don't contain 'https://'
Someone can confirm this?
Thanks,