Feature #1739
Improve monitorization
| Status: | Closed | Start date: | 01/25/2013 | |
|---|---|---|---|---|
| Priority: | Normal | Due date: | ||
| Assignee: |  Javi Fontan | % Done: | 0% | |
| Category: | Drivers - Auth | |||
| Target version: | Release 4.0 | |||
| Resolution: | fixed | Pull request: |
Description
1 - Reuse the monitorization to return all the information about all the VMs in the host, instead of an independent monitorization action per VM.
2 - Separe MONITOR commands from the rest so they can be ran simultaneously
Associated revisions
feature #1739: im values in new lines
fetaure #1739: make poll xen/kvm return all vms in template format
feature #1739: add vm poll scripts to xen and kvm im
feature #1739: change im vmware to report vms info
feature #1739: change ganglia probes to get all vms info
feature #1739: change ganglia probes to get all vms info
Revert "feature #1739: change ganglia probes to get all vms info"
Commit moved to branch feature-1739, eb82280d9290b640ef015ecba4b94b05d89cf607
This reverts commit ad9898fa36b3c69e9726aa9c61aa2d93d933207e.
feature #1739: IM driver returns information encoded in base64
feature #1739: EC2 information manager returns information encoded in base64
feature #1739: Dummy driver returns host information in base64 encoding
feature #1739: Moved SSL functions to one_util namespace. oned decodes messages from information drivers
Feature #1739: Minor code cleanup
Feature #1739: VectorAttribute::vector_value checks the value contents
Feature #1739: Refactor VMMDriver poll so it can be called from IMMDriver
Feature #1739: Fix individual VM Poll
Feature #1739: Update vm last_poll when the operation ends, instead of when it is triggered
Feature #1739: Do not set last_poll when a VM is deployed
Feature #1739: Detect missing VMs from the Host poll info, and move them to 'unknown'
Feature #1739: All VMs are set to unk if the host poll fails
Feature #1739: When an individual VM poll fails, the VM is set to unk
Feature #1739: Do not update poll info for VMs that should not be in the reported host
Feature #1739: Drivers only return VM/ID when the VM name is exactly one-id
Feature #1739: Move the logic of what to do with a failed poll to VMMDriver
Feature #1739: Drivers inform that the VM_POLL info is included, making things easier for the core when 0 vms are found
Feature #1739: Monitor disabled hosts also
Feature #1739: Improve host monitoring performance
We don't need to retrieve and parse each Host xml body column, we
already have the Hosts in the cache
Feature #1739: Monitor disabled hosts only if they have running VMs
Feature #1739: Update doxygen file
Feature #1739: Add messages when extra VMs are detected
feature #1739: Fix wrong host vmware polling when libvirt returns warnings
Feature #1739: fix VMMD log levels, and omit redundant monitoring info
Feature #1739: Decode the host poll info string also in case of error
Feature #1739: Remove unused var
Feature #1739: Touch remotes location at start time to force the refresh
Feature #1739: Move the host last_poll_time filtering to the DB query
Feature #1739: Merge oned.conf Host & VM monitoring intervals
Feature #1739: Change host monitoring timeout to 5 minutes
Feature #1739: Skip VMM monitoring the first period
Feature #1739: Clean Host error messages when the problem is solved
featue #1739: Move Host and VM update functions to each class.
featue #1739: Minor modifications
Feature #1739: Minor changes (code style, error message readability)
Feature #1739: If the VM poll info can't be parsed, the VM is set to 'unknown'
Feature #1739: Fix bug in VM monitoring parser
Feature #1739: Fix bug in VM monitoring parser
(cherry picked from commit 0277e7430a993b3c24e7496134566bf619f4142e)
History
#1
Updated by Javi Fontan over 8 years ago
- Status changed from New to Assigned
- Assignee changed from Jaime Melis to Javi Fontan
#2
Updated by Carlos Martín over 8 years ago
The hosts only have one type of template error message, monitorization, but the VMs can have different messages: attach failed, monitorization failed, etc.
As the code is now, if the VM has an operation message (e.g. could not attach disk), a successful monitorization will clear that message.
Maybe we could define two types of errors: the ones that inform of a failed operation, and the ones that can be resolved on their own and can be cleared (monitoring).
The failed operations could be stored in a vector attribute:
ERROR_MESSAGES = [ ERROR = "attach failed because...", ERROR = "shutdown failed because..." ]
#3
Updated by Ruben S. Montero over 8 years ago
- Status changed from Assigned to Closed
- Resolution set to fixed