forked from Nordstrom/vmstats
-
Notifications
You must be signed in to change notification settings - Fork 0
/
TODO.txt
35 lines (33 loc) · 2.23 KB
/
TODO.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Version 2.1
- Replace my Graphite UDP code with a network framework code (Most likely TCP for the time being) - done
- Write FakeCarbon client so you can test without a graphite install - done
- Change IP from 1.1.1.1 to 1_1_1_1 - sense & configure via flags - done
-> Modified to change everything in the hostname from . to _, since you'd never really want that (I don't think)
- Change host.example.com to host_example_com - done
Version 2.2
- Handle the 20 second internal statistics better
- Allow user to configure polling period - could be 1 or 5 minutes or whatever that's divisible / 20
- The user can configure the # of intervals to gather per polling period. So, you could configure to loop every 5
minutes with 15 20-second intervals (if you wanted to)
- Have the option to send raw the raw stats to graphite or average/mean the data
- Added internal statistics for how long it takes to do things and how many stats are created.
Version 2.3+
- make maven build FakeCarbon
- Send internal statistics to Graphite for graphing
- Handle carbon-cache going away - currently stats just stop getting pushed
- possibly cache 1 or 2 cycles of data and then die - in case the issue is intermittent
- Write code to compare stats coming out of vmstats and stats coming into FakeCarbon
- Get the parent information for a VM - so it becomes vmstats.CLUSTER.vmname.stat...
- A better/new manager thread that will help with exception handling
-> When an exception happens, the threads will be restarted instead of the entire app exiting (at least for
most exceptions)
- Figure out how to mock up the VMware web services so tests can run without an actual vCenter/ESX node
- With mock services, write unit tests
- Add back in sending to Graphite via UDP
- Add in sending to Graphite via AMQP
- Filter VM names better, so if someone has host*name(noreallytheotherone)&thathostname as a vm name, stats get sent.
- Carbon File Management
- Some sort of tool to manage the carbon files so that as some of these things get renamed, the data can be managed
in better ways
- Remove VM's that haven't been seen for X time
- Make a shell script to manage the startup to make the supervisord/upgrade process simple