On metrics with a lot of data, or on searches over longer time periods, the anomaly detection process will run out of allocated memory, tested up through a configured 4GB for the process. Need to optimize this process so that it works on smaller chunks of the data at a given time to better use server memory.

In most clock interfaces in the world, clicking the "Up" arrow actually decrements the hour/minute, but on StatusWolf the opposite happens

My webserver stack is CentOS 6.4 + Apache + php-fpm via mod_fastcgi.so. I've tried Ubuntu 12.04 also.

Everything seems to work up until I run a query for a graph, at which point the browser shows:

POST http://tsdbserver:8080/adhoc/search/OpenTSDB [HTTP/1.1 500 Internal Server Error 42ms]

In the php-fpm logs I see:

PHP Fatal error:  Call to undefined function apache_getenv() in /vm/prod/StatusWolf/lib/StatusWolf/Model/Datasource/OpenTSDB.php on line 277

Does StatusWolf depend on an Apache + mod_php setup? Thank you.

If I refresh an unsaved dashboard I lose it. But if I make changes to a search on a previously saved dashboard, are they... saved automagically? I tried clicking "Save Dashboard" after modifying a widget and it complained I already had a dashboard with that name, so not sure about the widget status.

One common use case is memory graphs - seeing each line unstacked makes less sense than seeing a flat line representing total resources at the top, and allocations as colored areas:

I keep seeing this with .73 of status wolf:

[Tue Oct 08 15:31:26 2013] [error] [client 10.131.252.254] PHP Parse error: syntax error, unexpected '[' in /var/www/StatusWolf/lib/StatusWolf/Model/Datasource/OpenTSDB.php on line 407, referer: https://10.131.253.226/dashboard/

Is it just me?

We need to be able to see all the hostnames and metric names bellow the graph without having to slide up and down with the mouse. Complex graphs can make the legend almost impossible to read.

Were using 0.74 of Status Wolf on an Ubuntu 12.04 server.

If we call for a metric everything works. When we pass in a tag and it's value I see the following in the log:

[Thu Oct 10 15:54:23 2013] [error] [client 10.135.15.248] PHP Notice: Undefined index: atmos.posts in /var/www/StatusWolf/lib/StatusWolf/Model/Datasource/OpenTSDB.php on line 437, referer: https://wolf.roccops.com/adhoc

Thanks,

I noticed this entry in the Apache error log:

PHP Notice: Undefined offset: 0 in /var/www/StatusWolf/lib/StatusWolf/Router/SWRouter.php on line 57

Which is this line:
if (count($url_parts['url_path'] == 1) && strlen($url_parts['url_path'][0]) == 0)

It has a mis-placed right parenthesis, and should be:
if (count($url_parts['url_path']) == 1 && strlen($url_parts['url_path'][0]) == 0)

Right now with tag substitution you can add a tag to all graphs. If a particular metric doesn't have that tag it will cause it not to load.

One suggestion may be that tags passed in on the command line only replace tags if they already exist in the metric and never add them to metrics where they previously didn't exist.

The use case we have is that we have a mysql dashboard where we want to sub host & port tags, but we also have the proc.stat.cpu.percent_used which is global to the server and doesn't have a port tag.

Hi,

This looks very promising, but some sort of installation document is needed. I figured out dependency on Pear/Auth, MDB2 and mysqli extension. However I can't find a way to add users. So I turned authentication off. But now I am stuck trying to figure out apache setup to work with request router that seems to be used here.

Limus

It would be very useful to be able to build a dashboard, but also load the dashboard and plug in specific parameters.

The classic example of this would be to build a set of graphs for a specific host but also be able to load that dashboard and pug in a different host.

In the same way you can override tags for a dashboard it would be helful to be able to override the time window with a range or a fixed period of time. That would allow us to build one dashboard, and then see the last day by default, but also "zoom" in to see the last hour only for all graphs. This would be helpful to put in links to the dashboard as well.

There are metrics where I would like to show data points more than once per minute.

Add Bart PHP library (https://github.com/box/bart) to StatusWolf to take advantage of its autoloading functionality

Useful when data points differ in scale by orders of magnitude.

I'm trying to create a graph, and it returns the following error:

Sorry but your request was rejected as being invalid.
The reason provided was:
Invalid downsampling specifier 'nointerpolation' in m=sum:nointerpolation:proc.loadavg.1min{host=DEV-TEST003}

Is our OpenTSDB version incompatible with StatusWolf v0.8.8?

Here's its manifest:

Specification-Title: OpenTSDB
Specification-Version: 1.1.0
Specification-Vendor: The OpenTSDB Authors
Implementation-Title: net.opentsdb
Implementation-Version: 6f1d105
Implementation-Vendor: The OpenTSDB Authors

And here are the StatusWolf debug log entries:

2014-01-09 10:02:32 - DEBUG --> (alan|192.168.100.102) OpenTSDB object created
2014-01-09 10:02:32 - DEBUG --> (alan|192.168.100.102) {"downsample_master_interval":"1","new_query":"true","title":"","period":"span-search","time_span":"14400","start_time":"1389276152","end_time":"1389290552","auto_update":"false","metrics":{"02d60f19bc88ffff6ce41e61c1227a0d":{"name":"proc.loadavg.1min","tags":["host=DEV-PROXY003"],"agg_type":"sum","ds_type":"sum","ds_interval":"1"}},"history_graph":"no","metrics_count":"1"}
2014-01-09 10:02:32 - ERROR --> (alan|192.168.100.102) Failed to retrieve metrics from OpenTSDB, start time was: 2014/01/09-06:02:32
2014-01-09 10:02:32 - ERROR --> (alan|192.168.100.102) Failed to fetch data from:
URL: http://localhost:4242/q?start=2014/01/09-06:02:32&end=2014/01/09-09:57:31&m=sum:nointerpolation:proc.loadavg.1min{host=DEV-TEST003}&ascii
Response Code: 400
Response Body: <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Trans

If I load the URL into a browser, and remove the "nointerpolation" keyword, it works:

http://localhost:4242/q?start=2014/01/09-06:02:32&end=2014/01/09-09:57:31&m=sum:proc.loadavg.1min{host=DEV-TEST003}&ascii

Also, if I enable the Interpolation option in the graph parameters form, it also works because in that case the keyword is removed from the query:

2014-01-09 10:47:07,022 INFO [New I/O worker #5] HttpQuery: [id: 0x623250a6, /127.0.0.1:59138 => /127.0.0.1:4242] HTTP /q?start=2014/01/09-06:47:07&end=2014/01/09-10:42:06&m=sum:proc.loadavg.1min{host=DEV-TEST003}&ascii done in 17ms

It looks like you're able to suggest a tag based on a metric however the suggestions for possible tag values don't seem to work.

From the opsnTSDB side when chooseing a tag = HTTP /suggest?type=tagk&q=hos
From the openTSDB side when chosing a tag value = HTTP /suggest?type=tagv&q=op

When using statuswolf = HTTP /suggest?type=metrics&q=proc.loadavg.1min+host%3D

I'll see if I can make a suggesting in the code and post an update. If there is a trick to getting this to work (autocompletion of possible tag values - please let me know)

Line 82 has http://opentsdb.ve.box.net:4242/suggest?type=metrics&q=

For some dashboards where I have a graph widget that is particularly "full", I'd like an easy way to just clone that widget in the current dashboard and then quickly edit it to reflect another metric. The current behavior requires me to manually add a new graph widget and then configure it's time interval, plotting options, and just about everything that I'd like to be the same as another graph on the dashboard, except for the metric to plot.

Initially would comprise the ability to manually annotate graphs and have the annotations saved so that when the metric is graphed again later the annotation will appear.

Must have an admin interface to manage users, saved searches, global config, etc.

Hi,

I came across statuswolf while trying to find a good way to display opentsdb data.

I followed the steps that you have mentioned, but I am not able to get anything on the web.

1. I followed instruction to install pear . based on your instructions.
2. I downloaded StatusWolf.tgz, un-tar`d it into my webroot which is /var/www/html.
3. Created mysql DB, created the auth.conf, datasource.conf and statuswolf.conf. I left everything to default.
4. I made the entry of the virtual host in /etc/httpd/conf/httpd.conf
5. Restarted the web service.

Now when I launch the web page (http://x.x.x.x/app/) and login in, i get options to create dashboard, view my settings & Ad-Hoc search. I clicked on create dashboard, and i get this error,

File does not exist: /var/www/html/StatusWolf-0.8.5/dashboard

So, I created this directory and did chmod -R 777 dashboard.

Now when I again click on create dashboard, nothing gets displayed but the blank dashboard directory.

I am missing something ?

sw.graphwidget.js:1976

widget.svg = d3.select('#' + graphdiv.attr('id')).append('svg')

editing / updating dashboards requires a lot of clicking. I can think of the following right now:

• Add new graph
• New ad-hoc search
• arrow keys to select graphs on dashboard
• e - edit current graph
• n - change current graph title
• u - toggle auto update
• a - add new metric to current graph
• t - change current graph time
• g - go ( save current graph changes)

In addition, I would suggest that when using keyboard shortcuts you could skip the transition animations to speed up the process.

Many problems are cyclic over the period of a day, having a 48 hour graph pulls less data than a week, and still lets you compare today to yesterday.

First series returned for the metric is assigned to the right axis, the rest are mapped to the left axis... Sample search params:

proc.stat.cpu.percentage_iowait, agg:avg, ds:avg on left axis
proc.stat.cpu.percentage_idle agg:avg, ds:avg on right axis

It would be nice to assign short names to dashboards and reference them by those names:

Something like:

server.com/StatusWolf/go/My+Dash

I followed all the steps as exactly mentioned in the Installation wiki page properly. I disabled the authentication.

While trying to open the dashboard and adhoc search page. I am getting the following errors:

File does not exist: /var/www/StatusWolf-0.5.2/dashboard, referer: http://opentsdb-1.*.com:9653/app/

File does not exist: /var/www/StatusWolf-0.5.2/adhoc, referer: http://opentsdb-1.*.com:9653/app/

Can someone please help me. Moreover i am getting errors like:

[Tue Sep 03 15:25:41 2013] [error] [client 172.17.81.236] PHP Notice: Undefined index: _sw_authsession in /var/www/StatusWolf-0.5.2/lib/StatusWolf/Router/SWRouter.php on line 26
[Tue Sep 03 15:25:41 2013] [error] [client 172.17.81.236] PHP Notice: Undefined index: _sw_authsession in /var/www/StatusWolf-0.5.2/lib/StatusWolf/Router/SWRouter.php on line 26
[Tue Sep 03 15:25:41 2013] [error] [client 172.17.81.236] PHP Notice: Undefined index: _sw_authsession in /var/www/StatusWolf-0.5.2/lib/StatusWolf/Views/header.php on line 21
[Tue Sep 03 15:25:41 2013] [error] [client 172.17.81.236] PHP Notice: Undefined index: _sw_authsession in /var/www/StatusWolf-0.5.2/lib/StatusWolf/Views/header.php on line 21

Right now you need to precisely hover over the line chart point in order to get its value.
It would be much easier if the label would be activated when mouse is moving horizontally on oX axis

We've done something similar in our own OpenTSDB frontend by using the etags attribute and picking out those etags from any metrics in the past hour.

def get_tag_suggestions(metric):
    response = requests.get(
        'http://127.0.0.1:4242/q?start=1h-ago&m=sum:rate:%s&json' % metric
    )

    response = json.loads(response.content)

    if 'etags' not in response:
        return []

    return response['etags'][0]

Planning on implementing this as another endpoint in ApiController.php similar to tsdb_metric_list.

It would be nice to have some sort of status or health end point that could be hit in statuswolf which would verify the functionality of mysql connection (if used) and opentsdb connection. Ideally I could point my monitoring system at our various statuswolf installations at say /health or /status or something and based on the return codes I could tell if statuswolf is ok or not.

OpenTSDB does not store units by default, so it would be useful to be able to set graph title/x axis / y axis labels for units, etc..

We need it to visualize data correctly on a huge set of hosts. For example try to display df.bytes.free of all hosts and you won't be able to visualize which hosts are really low on free disk space.

alternatively (or both) if the user has no dashboards add text like "No dashboards make your first search"

currently the graphs rendered in firefox v 25 look like this:

imho looks like some svg issue since the rect element has a reported width/height of 0px but i have no idea how to fix it

followed install steps but getting some errors and graphs are not working.

2013-12-24 11:25:16 - DEBUG --> (statuswolf_user|127.0.0.1) ["undefined","api","get_saved_searches"]
2013-12-24 11:25:16 - DEBUG --> (statuswolf_user|127.0.0.1) {"scheme":"http","host":"localhost","port":9653,"path":"/undefined/api/get_saved_searches","url_path":["undefined","api","get_saved_searches"]}
2013-12-24 11:25:16 - DEBUG --> Controller UndefinedController was not found, punting to Home
2013-12-24 11:25:16 - DEBUG --> [{"priority":7,"message":"AUTH: Auth::checkAuth() called."}]
2013-12-24 11:25:16 - DEBUG --> [{"priority":7,"message":"AUTH: Auth::checkAuth() called."},{"priority":6,"message":"AUTH: Session OK."}]
2013-12-24 11:25:16 - DEBUG --> [{"priority":7,"message":"AUTH: Auth::checkAuth() called."},{"priority":6,"message":"AUTH: Session OK."},{"priority":7,"message":"AUTH: Auth::start() called."}]
2013-12-24 11:25:16 - DEBUG --> [{"priority":7,"message":"AUTH: Auth::checkAuth() called."},{"priority":6,"message":"AUTH: Session OK."},{"priority":7,"message":"AUTH: Auth::start() called."},{"priority":7,"message":"AUTH: Auth::assignData() called."}]
2013-12-24 11:25:16 - DEBUG --> [{"priority":7,"message":"AUTH: Auth::checkAuth() called."},{"priority":6,"message":"AUTH: Session OK."},{"priority":7,"message":"AUTH: Auth::start() called."},{"priority":7,"message":"AUTH: Auth::assignData() called."},{"priority":7,"message":"AUTH: Auth::checkAuth() called."}]
2013-12-24 11:25:16 - DEBUG --> [{"priority":7,"message":"AUTH: Auth::checkAuth() called."},{"priority":6,"message":"AUTH: Session OK."},{"priority":7,"message":"AUTH: Auth::start() called."},{"priority":7,"message":"AUTH: Auth::assignData() called."},{"priority":7,"message":"AUTH: Auth::checkAuth() called."},{"priority":6,"message":"AUTH: Session OK."}]
2013-12-24 11:25:16 - DEBUG --> [{"priority":7,"message":"AUTH: Auth::checkAuth() called."},{"priority":6,"message":"AUTH: Session OK."},{"priority":7,"message":"AUTH: Auth::start() called."},{"priority":7,"message":"AUTH: Auth::assignData() called."},{"priority":7,"message":"AUTH: Auth::checkAuth() called."},{"priority":6,"message":"AUTH: Session OK."},{"priority":7,"message":"AUTH: Auth::checkAuth() called."}]
2013-12-24 11:25:16 - DEBUG --> [{"priority":7,"message":"AUTH: Auth::checkAuth() called."},{"priority":6,"message":"AUTH: Session OK."},{"priority":7,"message":"AUTH: Auth::start() called."},{"priority":7,"message":"AUTH: Auth::assignData() called."},{"priority":7,"message":"AUTH: Auth::checkAuth() called."},{"priority":6,"message":"AUTH: Session OK."},{"priority":7,"message":"AUTH: Auth::checkAuth() called."},{"priority":6,"message":"AUTH: Session OK."}]
2013-12-24 11:25:16 - DEBUG --> (statuswolf_user|127.0.0.1) ["undefined","app","js","lib","bootstrap-datetimepicker.js"]
2013-12-24 11:25:16 - DEBUG --> (statuswolf_user|127.0.0.1) {"scheme":"http","host":"localhost","port":9653,"path":"/undefined/app/js/lib/bootstrap-datetimepicker.js","url_path":["undefined","app","js","lib","bootstrap-datetimepicker.js"]}
2013-12-24 11:25:16 - DEBUG --> Controller UndefinedController was not found, punting to Home

-- http conf--
Listen 9653
<VirtualHost *:9653>
DocumentRoot /data/statuswolf/

Options FollowSymLinks
AllowOverride All

StatusWolf does not gracefully handle situations where OpenTSDB searches fail to return data, the interface hangs at "Parsing Metric Data". Need to implement full error handling for these conditions and alert the user that the OpenTSDB search has failed and they should try again.

via ssabo

Less work to re-create dashboards versus clicking in the UI.

Hello,

Thanks for developing this great GUI for OpenTSDB.

I'm encountering some php syntax error while creating my graphs via OpenTSDB datasource.

Here is the error :

PHP Parse error: syntax error, unexpected '[' in /WEBS/opentsdb/opentsdb.priv.atos.fr/docs/lib/StatusWolf/Model/Datasource/OpenTSDB.php on line 339, referer: http://opentsdb.priv.atos.fr:9653/adhoc

The corresponding line is :

$search_key = (array_keys($query_bits['metrics'])[0]);

Any help will be greatly appreciated :) !

Regards,

François

it's unintuitive that your dashboard goes away without prompting to save it

Allows users to manipulate graphs quickly and with less clicking.

When building a graph you should be able to add a threshold, or a low and a high threshold, which will display on the graph as an indicator of max/min expected values.

Requested by Kunal

Just a little .sql file would do.

I got StatusWolf working with no errors in apache log. Thank you for the installation document. Need to add that php needs to be compiled with --with-curl and --with-mysqli options.

However, ad-hoc query screen looks pretty garbled. I am attaching image.

Firefox 22 on Windows. As far as I can tell all stylesheet and javascript files are downloaded. I am going to dig some more with firebug but in the meantime any suggestions would be welcome. Below is access log from apache:

10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /adhoc/ HTTP/1.1" 200 17662
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/css/statuswolf_base.css?v=1 HTTP/1.1" 200 14211
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/css/nav.css?v=1 HTTP/1.1" 200 419
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/css/adhoc.css?v=1.0 HTTP/1.1" 200 3811
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/css/popups.css?v=1.0 HTTP/1.1" 200 2673
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/js/sw_lib.js HTTP/1.1" 200 743
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/js/status_wolf_colors.js HTTP/1.1" 200 7539
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/js/lib/date.js HTTP/1.1" 200 25809
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/js/lib/magnific-popup.js HTTP/1.1" 200 23772
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/js/lib/bootstrap.js HTTP/1.1" 200 61764
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/js/lib/dygraph-combined.js HTTP/1.1" 200 135365
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/js/lib/jquery.js HTTP/1.1" 200 92556
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "POST /api/get_saved_searches HTTP/1.1" 200 2
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /favicon.ico HTTP/1.1" 200 70397
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /api/datasource_form/opentsdb_search HTTP/1.1" 200 67343
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/css/datetimepicker.css HTTP/1.1" 200 3208
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/css/toggle-buttons.css HTTP/1.1" 200 1177
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/css/push-button.css HTTP/1.1" 200 862
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/css/table.css HTTP/1.1" 200 487
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/css/tooltip.css HTTP/1.1" 200 1476
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/css/loader.css HTTP/1.1" 200 2122
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/js/lib/bootstrap-datetimepicker.js HTTP/1.1" 200 47138
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/js/toggle-buttons.js HTTP/1.1" 200 413
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/js/push-button.js HTTP/1.1" 200 759
10.180.4.250 - - [30/Jul/2013:09:12:48 -0700] "GET /app/js/lib/jquery.autocomplete.js HTTP/1.1" 200 20549

Limus

It's helpful to know when code changed. We'd like to have our build system annotate graphs when deployments succeed.

Our list of saved dashboards is getting unmanageable. Some ways to help sort through them would be nice.

suggestions:

1. Sort options (Alphabetical, Chronological from creation time, Owner, etc)
2. A search bar (search for user, shared name, or metics used)

https://s3.amazonaws.com/uploads.hipchat.com/23947/185789/qaxw6gr8p7mv329/error.png

Just an FYI. When rendering graphs, receive this error message with very little details running against OpenTSDB 2.0. Does not happen against 1.0 of course.