Some extra numbers, to explore the lower limits. The A10/Neutron CI uses the following VMs:
Jenkins/zuul/gearman, Ubuntu 14.04, 1 vcpu, 1GB ram (t2.micro) Devstack-gate, Ubuntu 14.04, 1 vcpu, 4GB ram (t2.medium) Note that 4GB doesn’t work if you enable swift; that needs 8gb and more cores. Since we run a subset of neutron, our test time is 18 minutes. I need three devstack-gate slaves to keep up with the gate crush, or about 9-10/hour total throughput. One is plenty with the current level of submissions. Thanks, doug From: Amit Das <[email protected]<mailto:[email protected]>> Date: Friday, October 31, 2014 at 7:15 AM To: Erlon Cruz <[email protected]<mailto:[email protected]>> Cc: Marcus Vinícius Ramires do Nascimento <[email protected]<mailto:[email protected]>>, "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Subject: Re: [OpenStack-Infra] [openstack-infra] [third-party CI] Hardware requirements for setting up CI environment Thanks a lot Erlon. The document looks really good. Will update how our installation goes through. Regards, Amit CloudByte Inc.<http://www.cloudbyte.com/> On Fri, Oct 31, 2014 at 6:25 PM, Erlon Cruz <[email protected]<mailto:[email protected]>> wrote: Hi Amit, Bharat Attached the tutorial. On Fri, Oct 31, 2014 at 10:15 AM, Amit Das <[email protected]<mailto:[email protected]>> wrote: Thanks Erlon & Jeremy for the detailed responses. Erlon, the configuration doc will definitely help us. Looking forward to it. Regards, Amit CloudByte Inc.<http://www.cloudbyte.com/> On Fri, Oct 31, 2014 at 5:29 PM, Erlon Cruz <[email protected]<mailto:[email protected]>> wrote: We have a CI for Cinder drivers. Currently we use 1 instance for the master and 1 instance for each slave(we test one driver on each slave). The configuration is the following: Master: 2VCPUs + 8GB RAM Slaves: 2VCPUs + 12GB RAM Network: 4Mb Download + 4Mb Upload Our jobs usually take 1:20h to run, but that is because of the network. In our setup we virtualize all hosts using VMWare, which is very handy as we can save an snapshot and trigger a revert after the job is complete also, if you have FC drivers, VMWare will save a lot of pain with it PCI passthrought. We have an step by step to configure the env with VMWare and Cinder drivers, if you need we can make it available. On Thu, Oct 30, 2014 at 1:24 PM, Jeremy Stanley <[email protected]<mailto:[email protected]>> wrote: On 2014-10-30 20:28:17 +0530 (+0530), Amit Das wrote: > This is exactly what we intend to do. Wanted to know if there is > anything else that we should know before placing the requests. In that case, if you're wanting to do devstack-gate based jobs we use instances with 8 VCPUs and 8 GB RAM. A typical tempest-full run takes somewhere in the vicinity of an hour to complete depending on what configuration options have been selected. Due to the inability to cleanly re-run DevStack or to trust that proposed changes haven't destroyed the usability of the worker instance, we recycle and replace them with fresh instances for each job. -- Jeremy Stanley _______________________________________________ OpenStack-Infra mailing list [email protected]<mailto:[email protected]> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-infra
_______________________________________________ OpenStack-Infra mailing list [email protected] http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-infra
