Re: [Openstack-operators] [gnocchi] monitoring storage use case inquiry
thanks for the feedback. the original reason for dropping it was because there was no active development on the driver for many months and it remained broken (functionality and from testing pov) so rather than mislead people and say the driver worked, we thought it better to drop it. if someone is willing to support the driver i believe the community would gladly add it back in, but to be honest, we haven't had any requests for it. as we are a small group, we decided to focus our efforts on what we know best and the driver that we know people are using. (full disclosure: i use Ceph driver). cheers, -- gord From: Erik McCormick <emccorm...@cirrusseven.com> Sent: August 1, 2016 12:52:10 PM To: Sam Morrison Cc: gordon chung; openstack-operators Subject: Re: [Openstack-operators] [gnocchi] monitoring storage use case inquiry On Jul 31, 2016 8:32 PM, "Sam Morrison" <sorri...@gmail.com<mailto:sorri...@gmail.com>> wrote: > > Hi Gordon, > > We are using the influxDB backend and we have our retention policies set to: > > Every minute for an hour > Every 10 minutes for a day > Every hour for a year > > Currently we hover around 8,000 instances. > > We understand the influxDB driver was taken out of gnocchi, bit annoyed as it > wasn’t mentioned on the operators list. > We currently have just got it working with version 2.1 of gnocchi and are > keen to see if it can be added back into gnocchi. > > We already use influxDB for other non openstack stuff and so would rather use > that as opposed to adding yet another system. > > Would be interested to know what other operators use gnocchi and what backend > they use. > I haven't implemented Gnocchi yet, but it's on my to do list. We are using Influxdb for our collectd data and had planned to use that for Gnocchi as well. +1 for putting the driver back in as I didn't know it was gone :(. -Erik > Cheers, > Sam > > > > On 29 Jul 2016, at 11:30 PM, gordon chung > > <g...@live.ca<mailto:g...@live.ca>> wrote: > > > > hi folks, > > > > the Gnocchi dev team is working on pushing out a new serialization > > format to improve disk footprint and while we're at it, we're looking at > > other changes as well. to get a bit more insight to help decide what > > changes we make, one useful metric would be to know what your > > requirements are for storing data. as you may know Gnocchi does not > > store raw datapoints but aggregates data to a specified granularity (eg. > > 5s, 30s, 1min, 1 day, etc...). what we're after is what's the longest > > timeseries you're capturing or hoping to capture? a datapoint every > > minute for a day/week/month/year? a datapoint every 10mins for a > > week/month/year? something else? > > > > your feedback would be greatly appreciated. > > > > cheers, > > > > -- > > gord > > ___ > > OpenStack-operators mailing list > > OpenStack-operators@lists.openstack.org<mailto:OpenStack-operators@lists.openstack.org> > > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators > > > ___ > OpenStack-operators mailing list > OpenStack-operators@lists.openstack.org<mailto:OpenStack-operators@lists.openstack.org> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
Re: [Openstack-operators] [gnocchi] monitoring storage use case inquiry
On Jul 31, 2016 8:32 PM, "Sam Morrison"wrote: > > Hi Gordon, > > We are using the influxDB backend and we have our retention policies set to: > > Every minute for an hour > Every 10 minutes for a day > Every hour for a year > > Currently we hover around 8,000 instances. > > We understand the influxDB driver was taken out of gnocchi, bit annoyed as it wasn’t mentioned on the operators list. > We currently have just got it working with version 2.1 of gnocchi and are keen to see if it can be added back into gnocchi. > > We already use influxDB for other non openstack stuff and so would rather use that as opposed to adding yet another system. > > Would be interested to know what other operators use gnocchi and what backend they use. > I haven't implemented Gnocchi yet, but it's on my to do list. We are using Influxdb for our collectd data and had planned to use that for Gnocchi as well. +1 for putting the driver back in as I didn't know it was gone :(. -Erik > Cheers, > Sam > > > > On 29 Jul 2016, at 11:30 PM, gordon chung wrote: > > > > hi folks, > > > > the Gnocchi dev team is working on pushing out a new serialization > > format to improve disk footprint and while we're at it, we're looking at > > other changes as well. to get a bit more insight to help decide what > > changes we make, one useful metric would be to know what your > > requirements are for storing data. as you may know Gnocchi does not > > store raw datapoints but aggregates data to a specified granularity (eg. > > 5s, 30s, 1min, 1 day, etc...). what we're after is what's the longest > > timeseries you're capturing or hoping to capture? a datapoint every > > minute for a day/week/month/year? a datapoint every 10mins for a > > week/month/year? something else? > > > > your feedback would be greatly appreciated. > > > > cheers, > > > > -- > > gord > > ___ > > OpenStack-operators mailing list > > OpenStack-operators@lists.openstack.org > > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators > > > ___ > OpenStack-operators mailing list > OpenStack-operators@lists.openstack.org > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
Re: [Openstack-operators] [gnocchi] monitoring storage use case inquiry
Hi Gordon, We are using the influxDB backend and we have our retention policies set to: Every minute for an hour Every 10 minutes for a day Every hour for a year Currently we hover around 8,000 instances. We understand the influxDB driver was taken out of gnocchi, bit annoyed as it wasn’t mentioned on the operators list. We currently have just got it working with version 2.1 of gnocchi and are keen to see if it can be added back into gnocchi. We already use influxDB for other non openstack stuff and so would rather use that as opposed to adding yet another system. Would be interested to know what other operators use gnocchi and what backend they use. Cheers, Sam > On 29 Jul 2016, at 11:30 PM, gordon chungwrote: > > hi folks, > > the Gnocchi dev team is working on pushing out a new serialization > format to improve disk footprint and while we're at it, we're looking at > other changes as well. to get a bit more insight to help decide what > changes we make, one useful metric would be to know what your > requirements are for storing data. as you may know Gnocchi does not > store raw datapoints but aggregates data to a specified granularity (eg. > 5s, 30s, 1min, 1 day, etc...). what we're after is what's the longest > timeseries you're capturing or hoping to capture? a datapoint every > minute for a day/week/month/year? a datapoint every 10mins for a > week/month/year? something else? > > your feedback would be greatly appreciated. > > cheers, > > -- > gord > ___ > OpenStack-operators mailing list > OpenStack-operators@lists.openstack.org > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
[Openstack-operators] [gnocchi] monitoring storage use case inquiry
hi folks, the Gnocchi dev team is working on pushing out a new serialization format to improve disk footprint and while we're at it, we're looking at other changes as well. to get a bit more insight to help decide what changes we make, one useful metric would be to know what your requirements are for storing data. as you may know Gnocchi does not store raw datapoints but aggregates data to a specified granularity (eg. 5s, 30s, 1min, 1 day, etc...). what we're after is what's the longest timeseries you're capturing or hoping to capture? a datapoint every minute for a day/week/month/year? a datapoint every 10mins for a week/month/year? something else? your feedback would be greatly appreciated. cheers, -- gord ___ OpenStack-operators mailing list OpenStack-operators@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators