[vdsm] exploring a possible integration between collectd and Vdsm
Hi all, In the last 2.5 days I was exploring if and how we can integrate collectd and Vdsm. The final picture could look like: 1. collectd does all the monitoring and reporting currently Vdsm does 2. Engine consumes data from collectd 3. Vdsm consumes *notifications* from collectd - for few but important tasks like Drive high water mark monitoring Benefits (aka: why to bother?): 1. less code in Vdsm / long-awaited modularization of Vdsm 2. better integration with the system, reuse of well-known components 3. more flexibility in monitoring/reporting: collectd is special purpose existing solution 4. faster, more scalable operation because all the monitoring can be done in C At first glance, Collectd seems to have all the tools we need. 1. A plugin interface (https://collectd.org/wiki/index.php/Plugin_architecture and https://collectd.org/wiki/index.php/Table_of_Plugins) 2. Support for notifications and thresholds (https://collectd.org/wiki/index.php/Notifications_and_thresholds) 3. a libvirt plugin https://collectd.org/wiki/index.php/Plugin:virt So, the picture is like 1. we start requiring collectd as dependency of Vdsm 2. we either configure it appropriately (collectd support config drop-ins: /etc/collectd.d) or we document our requirements (or both) 3. collectd monitors the hosts and libvirt 4. Engine polls collectd 5. Vdsm listens from notifications Should libvirt deliver us the event we need (see https://bugzilla.redhat.com/show_bug.cgi?id=1181659), we can just stop using collectd notifications, everything else works as previously. Challenges: 1. Collectd does NOT consider the plugin API stable (https://collectd.org/wiki/index.php/Plugin_architecture#The_interface.27s_st...) so the plugins should be inclueded in the main tree, much like the modules of the linux kernel Worth mentioning that the plugin API itself has a good deal of rough edges. we will need to maintain this plugin ourselves, *and* we need to maintain our thin API layer, to make sure the plugin loads and works with recent versions of collectd. 2. the virt plugin is out of date, doesn't report some data we need: see https://github.com/collectd/collectd/issues/1945 3. the notification message(s) are tailored for human consumption, those messages are not easy to parse for machines. 4. the threshold support in collectd seems to match values against constants; it doesn't seem possible to match a value against another one, as we need to do for high water monitoring (capacity VS allocation). How I'm addressing, or how I plan to address those challenges (aka action items): 1. I've been experimenting with out-of-tree plugins, and I managed develop, build, install and run one out-of-tree plugin: https://github.com/mojaves/vmon/tree/master/collectd The development pace of collectd looks sustainable, so this doesn't look such a big deal. Furthermore, we can engage with upstream to merge our plugins, either as-is or to extend existing ones. 2. Write another collectd plugin based on the Vdsm python code and/or my past accelerator executable project (https://github.com/mojaves/vmon) 3. patch the collectd notification code. It is yet another plugin OR 4. send notification from the new virt module as per #2, bypassing the threshold system. This move could preclude the new virt module to be merged in the collectd tree. Current status of the action items: 1. done BUT PoC quality 2. To be done (more work than #1/possible dupe with github issue) 3. need more investigation, conflicts with #4 4. need more investigation, conflicts with #3 All the code I'm working on will be found on https://github.com/mojaves/vmon Comments are appreciated -- Francesco Romani RedHat Engineering Virtualization R & D Phone: 8261328 IRC: fromani
On Tue, Oct 11, 2016 at 2:05 PM, Francesco Romani <fromani@redhat.com> wrote:
Hi all,
In the last 2.5 days I was exploring if and how we can integrate collectd and Vdsm.
Some comments regarding storage high watermarks only. I will comment later on other aspects.
The final picture could look like: 1. collectd does all the monitoring and reporting currently Vdsm does 2. Engine consumes data from collectd 3. Vdsm consumes *notifications* from collectd - for few but important tasks like Drive high water mark monitoring
Drive high watermark is our core business, we cannot outsource it to collectd. Vdsm will always monitor high watermarks directly from libvirt.
Benefits (aka: why to bother?): 1. less code in Vdsm / long-awaited modularization of Vdsm 2. better integration with the system, reuse of well-known components 3. more flexibility in monitoring/reporting: collectd is special purpose existing solution 4. faster, more scalable operation because all the monitoring can be done in C
If the problem in monitoring is python, we can have small and simple helper doing the monitoring (for storage), like ioprocess.
At first glance, Collectd seems to have all the tools we need. 1. A plugin interface (https://collectd.org/wiki/index.php/Plugin_architecture and https://collectd.org/wiki/index.php/Table_of_Plugins) 2. Support for notifications and thresholds (https://collectd.org/wiki/index.php/Notifications_and_thresholds)
Setting threshhold and getting notifications when treshold is reached sounds like the best design for monitoring drive high watermarks. But I would like to depend on component that does *only* this task, and service only vdsm.
3. a libvirt plugin https://collectd.org/wiki/index.php/Plugin:virt
So, the picture is like
1. we start requiring collectd as dependency of Vdsm 2. we either configure it appropriately (collectd support config drop-ins: /etc/collectd.d) or we document our requirements (or both) 3. collectd monitors the hosts and libvirt 4. Engine polls collectd 5. Vdsm listens from notifications
Sounds good
Should libvirt deliver us the event we need (see https://bugzilla.redhat.com/show_bug.cgi?id=1181659), we can just stop using collectd notifications, everything else works as previously.
Challenges: 1. Collectd does NOT consider the plugin API stable (https://collectd.org/wiki/index.php/Plugin_architecture#The_interface.27s_st...) so the plugins should be inclueded in the main tree, much like the modules of the linux kernel Worth mentioning that the plugin API itself has a good deal of rough edges. we will need to maintain this plugin ourselves, *and* we need to maintain our thin API layer, to make sure the plugin loads and works with recent versions of collectd. 2. the virt plugin is out of date, doesn't report some data we need: see https://github.com/collectd/collectd/issues/1945 3. the notification message(s) are tailored for human consumption, those messages are not easy to parse for machines. 4. the threshold support in collectd seems to match values against constants; it doesn't seem possible to match a value against another one, as we need to do for high water monitoring (capacity VS allocation).
How I'm addressing, or how I plan to address those challenges (aka action items): 1. I've been experimenting with out-of-tree plugins, and I managed develop, build, install and run one out-of-tree plugin: https://github.com/mojaves/vmon/tree/master/collectd The development pace of collectd looks sustainable, so this doesn't look such a big deal. Furthermore, we can engage with upstream to merge our plugins, either as-is or to extend existing ones. 2. Write another collectd plugin based on the Vdsm python code and/or my past accelerator executable project (https://github.com/mojaves/vmon) 3. patch the collectd notification code. It is yet another plugin OR 4. send notification from the new virt module as per #2, bypassing the threshold system. This move could preclude the new virt module to be merged in the collectd tree.
Current status of the action items: 1. done BUT PoC quality 2. To be done (more work than #1/possible dupe with github issue) 3. need more investigation, conflicts with #4 4. need more investigation, conflicts with #3
All the code I'm working on will be found on https://github.com/mojaves/vmon
Comments are appreciated
-- Francesco Romani RedHat Engineering Virtualization R & D Phone: 8261328 IRC: fromani
On Tue, Oct 11, 2016 at 2:05 PM, Francesco Romani <fromani@redhat.com> wrote:
Hi all,
In the last 2.5 days I was exploring if and how we can integrate collectd and Vdsm.
The final picture could look like: 1. collectd does all the monitoring and reporting currently Vdsm does 2. Engine consumes data from collectd 3. Vdsm consumes *notifications* from collectd - for few but important tasks like Drive high water mark monitoring
Benefits (aka: why to bother?): 1. less code in Vdsm / long-awaited modularization of Vdsm 2. better integration with the system, reuse of well-known components 3. more flexibility in monitoring/reporting: collectd is special purpose existing solution 4. faster, more scalable operation because all the monitoring can be done in C
At first glance, Collectd seems to have all the tools we need. 1. A plugin interface (https://collectd.org/wiki/ index.php/Plugin_architecture and https://collectd.org/wiki/ index.php/Table_of_Plugins) 2. Support for notifications and thresholds (https://collectd.org/wiki/ index.php/Notifications_and_thresholds) 3. a libvirt plugin https://collectd.org/wiki/index.php/Plugin:virt
So, the picture is like
1. we start requiring collectd as dependency of Vdsm 2. we either configure it appropriately (collectd support config drop-ins: /etc/collectd.d) or we document our requirements (or both) 3. collectd monitors the hosts and libvirt 4. Engine polls collectd 5. Vdsm listens from notifications
Should libvirt deliver us the event we need (see https://bugzilla.redhat.com/show_bug.cgi?id=1181659), we can just stop using collectd notifications, everything else works as previously.
Challenges: 1. Collectd does NOT consider the plugin API stable ( https://collectd.org/wiki/index.php/Plugin_architecture# The_interface.27s_stability) so the plugins should be inclueded in the main tree, much like the modules of the linux kernel Worth mentioning that the plugin API itself has a good deal of rough edges. we will need to maintain this plugin ourselves, *and* we need to maintain our thin API layer, to make sure the plugin loads and works with recent versions of collectd. 2. the virt plugin is out of date, doesn't report some data we need: see https://github.com/collectd/collectd/issues/1945 3. the notification message(s) are tailored for human consumption, those messages are not easy to parse for machines. 4. the threshold support in collectd seems to match values against constants; it doesn't seem possible to match a value against another one, as we need to do for high water monitoring (capacity VS allocation).
How I'm addressing, or how I plan to address those challenges (aka action items): 1. I've been experimenting with out-of-tree plugins, and I managed develop, build, install and run one out-of-tree plugin: https://github.com/mojaves/ vmon/tree/master/collectd The development pace of collectd looks sustainable, so this doesn't look such a big deal. Furthermore, we can engage with upstream to merge our plugins, either as-is or to extend existing ones. 2. Write another collectd plugin based on the Vdsm python code and/or my past accelerator executable project (https://github.com/mojaves/vmon) 3. patch the collectd notification code. It is yet another plugin OR 4. send notification from the new virt module as per #2, bypassing the threshold system. This move could preclude the new virt module to be merged in the collectd tree.
Current status of the action items: 1. done BUT PoC quality 2. To be done (more work than #1/possible dupe with github issue) 3. need more investigation, conflicts with #4 4. need more investigation, conflicts with #3
All the code I'm working on will be found on https://github.com/mojaves/ vmon
Comments are appreciated
This generally sounds like a good idea - and I hope it is coordinated with our efforts for monitoring (see [1], [2]). Note that ages ago, ovirt-node actually had it already[3]. Few notes: - I think the most compelling reason to move to collectd is actually to benefit from the already existing plugins that it already has, which will cover a lot of the missing monitoring requirements and wishes we have (example: local disk usage on the host), as well as integrate it into Engine monitoring (example: postgresql performance monitoring). - You can't remove monitoring from VDSM - as it new VDSM may work against older Engine setups. You can gradually remove them. I'd actually begin with cleanup - there are some 'metrics' that are simply not needed and should not be reported in the first place and are there for historic reasons only. Remove them - from Engine first, from the DB and all, then later we can either send fake values or remove from VDSM. - If you are moving to collectd, as you can see from the metrics effort, we'd really want to send it elsewhere - and Engine should consume it from there. Metrics storages usually have a very nice REST UI with the ability to bring series with average, with different criteria (such as per hour, per minute or what not stats), etc. - I agree with Nir about separating between our core business and the monitoring we do for extra. Keep in mind that some of the stats are for SLA and critical scheduling decisions as well. - The libvirt collectd plugin is REALLY outdated. I think it may require significant work to bring it up to speed with our existing capabilities. Y. [1] https://sradcoblog.wordpress.com/2016/07/19/ovirt-metrics-elk/ [2] https://bronhaim.wordpress.com/2016/06/26/ovirt-metrics/ [3] https://github.com/oVirt/ovirt-node/blob/master/scripts/collectd.conf.in
-- Francesco Romani RedHat Engineering Virtualization R & D Phone: 8261328 IRC: fromani _______________________________________________ Devel mailing list Devel@ovirt.org http://lists.ovirt.org/mailman/listinfo/devel
----- Original Message -----
From: "Yaniv Kaul" <ykaul@redhat.com> To: "Francesco Romani" <fromani@redhat.com> Cc: "devel" <devel@ovirt.org> Sent: Tuesday, October 11, 2016 10:31:14 PM Subject: Re: [ovirt-devel] [vdsm] exploring a possible integration between collectd and Vdsm
On Tue, Oct 11, 2016 at 2:05 PM, Francesco Romani <fromani@redhat.com> wrote:
Hi all,
In the last 2.5 days I was exploring if and how we can integrate collectd and Vdsm. [...] This generally sounds like a good idea - and I hope it is coordinated with our efforts for monitoring (see [1], [2]).
Sure it will. I played a couple of days with collectd to "just" see a. how hard is to write a collectd plugin, and/or if it is feasible to ship ot out-of-tree for the initial few releases, until it stabilizes so it can be submitted upstream b. if we can get events/notifications from collectd c. if we can integrate those notifications with Vdsm And turns out we *can* do all of the above, with various degrees of difficulty.
Few notes: - I think the most compelling reason to move to collectd is actually to benefit from the already existing plugins that it already has, which will cover a lot of the missing monitoring requirements and wishes we have (example: local disk usage on the host), as well as integrate it into Engine monitoring (example: postgresql performance monitoring).
Agreed
- You can't remove monitoring from VDSM - as it new VDSM may work against older Engine setups. You can gradually remove them.
Yes, for example we can make Vdsm poll collectd and act as facade to old Engines, while new one should skip this step and ask collectd or the metrics aggregator service you mention below.
I'd actually begin with cleanup - there are some 'metrics' that are simply not needed and should not be reported in the first place and are there for historic reasons only. Remove them - from Engine first, from the DB and all, then later we can either send fake values or remove from VDSM.
Yes, this is the first place where we need to coordinate with the metrics effort.
- If you are moving to collectd, as you can see from the metrics effort, we'd really want to send it elsewhere - and Engine should consume it from there. Metrics storages usually have a very nice REST UI with the ability to bring series with average, with different criteria (such as per hour, per minute or what not stats), etc.
Fully agreed
- I agree with Nir about separating between our core business and the monitoring we do for extra. Keep in mind that some of the stats are for SLA and critical scheduling decisions as well.
Yes, of course adding a dependency for core monitoring is risky. So far the bottom line is that relying on collectd for this is just one more option on the table now. [mostly brainstorming from now on] However, I'd like highlight that is not just risky: is a different tradeoff. Doing the core monitoring in Vdsm (so in python, essentially in a single threaded server) is not a free lunch, because this has a quite high price on performance level. If the main Vdsm process is overloaded, then the polling cycle can get longer, and the overall response time of processing system events (e.g. disk detected full) can get longer as well. We've observed in not-so-distant past high response time from heavily loaded Vdsm. I think the idea of having different instances for different monitoring purposes (credit to Nir) is the best shot at the moment. We could maybe have one standard system collectd for regular monitoring, and perhaps one special purpose, very limited collectd instance for critical information. On top of that, Vdsm could double-checl and keep doing the core monitoring itself, albeit at lower rate (e.g. every 10s instead of every 2s; every 60s instead of every 15s). Leveraging libvirt events is *the* right thing, no doubt about that, but it would be very nice to have a dependable external service which can generate the events we need based on libvirt data, and move the notification logic on top of it. Something like (final picture, excluding intermediate compatibility layers) [data source] -----+------- | `-> [monitoring/metrics collection] -------------+--------------- | +--> [metrics store] -{data}-> [Engine] | `--> [notification service] -{events}-> [Vdsm] Not all the "boxes" need to be separate processes, for example collectd has some support for thresholds and notifications which is ~80% of what Vdsm needs (again not considering reliability, just feature-wise). [end brainstorm]
- The libvirt collectd plugin is REALLY outdated. I think it may require significant work to bring it up to speed with our existing capabilities.
Yep I looked briefly at that code. It is REALLY outdated :) Besides some information just not reported (easy to fix), we will most likely need to have here some logic to deal with stuck VMs, much like we do in Vdsm. In the past I ported some Vdsm monitoring code to C[1], this could perhaps help now. Bests, +++ [1] https://github.com/mojaves/vmon/tree/master/src -- Francesco Romani Red Hat Engineering Virtualization R & D Phone: 8261328 IRC: fromani
----- Original Message -----
From: "Francesco Romani" <fromani@redhat.com> To: "devel" <devel@ovirt.org> Sent: Wednesday, October 12, 2016 9:29:37 AM Subject: Re: [ovirt-devel] [vdsm] exploring a possible integration between collectd and Vdsm
----- Original Message -----
From: "Yaniv Kaul" <ykaul@redhat.com> To: "Francesco Romani" <fromani@redhat.com> Cc: "devel" <devel@ovirt.org> Sent: Tuesday, October 11, 2016 10:31:14 PM Subject: Re: [ovirt-devel] [vdsm] exploring a possible integration between collectd and Vdsm
On Tue, Oct 11, 2016 at 2:05 PM, Francesco Romani <fromani@redhat.com> wrote:
Hi all,
In the last 2.5 days I was exploring if and how we can integrate collectd and Vdsm. [...] This generally sounds like a good idea - and I hope it is coordinated with our efforts for monitoring (see [1], [2]).
Sure it will.
Sure it will be coordinated with the monitoring efforts*. -- Francesco Romani Red Hat Engineering Virtualization R & D Phone: 8261328 IRC: fromani
participants (3)
-
Francesco Romani -
Nir Soffer -
Yaniv Kaul