summaryrefslogtreecommitdiffstats
path: root/bifrost
AgeCommit message (Collapse)AuthorFilesLines
2017-08-11prototypes: bifrost: Fix ELEMENTS_PATH for latest diskimage-builder releaseMarkos Chandras2-2/+2
Upcoming diskimage-builder version (2.0.0) changed the location on where dib elements are installed from /usr/share/diskimage-builder to /usr/lib/$python/diskimage_builder/. However, base elements are included by default so fix the script to set ELEMENTS_PATH to the path of our out-of-tree elements. Change-Id: Ib5c4e1465a67a819c9d495b0e0a64addeb0294cb Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11prototypes: bifrost: Switch to JSON for inventoryMarkos Chandras1-2/+4
Upstream commit 3aaed64e88e3 ("Use JSON baremetal data in testing") switched from CSV to JSON for the inventory file in order to support the *_ipmitool drivers and virtualbmc. This commit mimics the upstream one. Depends-On: Ib5c4e1465a67a819c9d495b0e0a64addeb0294cb Change-Id: If66810db445cf997c5eb61090815abe186e002fa Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11prototypes: bifrost: Collect logs even if deployment failsMarkos Chandras1-2/+5
Move the collect-test-info.sh script to the 'exit' trap so we can collect whatever logs we can even if the bifrost deployment failed. Upstream change: https://review.openstack.org/433532 Change-Id: Icfbec01d894ff33e83730a40a09dc5540008dbaa Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11bifrost: Destroy all VMs from the slave nodesMarkos Chandras1-6/+6
New patchsets may introduce (or remove) testing VMs so make sure the script destroys all of them Change-Id: Ic96142823b9a8b044306327ce3454fbdc6fa5eb4 Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11prototypes: bifrost: Cleanup old VM console logsMarkos Chandras1-1/+1
The VM logs may not be rotated by logrotate so they are accumulated over time. Make sure they are gone when we start a new job. Change-Id: Iecf7e53015740db98c17a42612da7f44f3c4a676 Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11prototypes: bifrost: Drop the entire ironic databaseMarkos Chandras1-5/+2
There is no reason to only drop a few tables from it, so just drop the entire database. Change-Id: Iea9757fc5ccd85bc223d3c86d6a4c3ecf65fa130 Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11bifrost: Use 'unsafe' disk cache for test VMsMarkos Chandras1-0/+1
The VMs which are created by bifrost are not meant to be used in production so improve the overall CI job performance by switching to the 'unsafe' disk cache mode. Make it also the default mode when using bifrost outside of the Jenkins job. Change-Id: If88f6c9395617a4aa860f8d00e22abfdd18d2922 Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11jjb: infra: prototypes: bifrost: Ignore mysql and ironic failuresMarkos Chandras1-8/+10
On brand new hosts ironic may not be installed yet so attemping to restart ironic services will lead to failures. Similarly, mysql may also not be installed so do not try to wipe the tables. Change-Id: Ic77d9358aeecb11f5c076a47cfea18ee4f0dff9f Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11prototypes: bifrost: Increase default provision timeout to 3600 secondsMarkos Chandras1-1/+1
Sometimes, the current '2400 seconds' timeout is not enough to all 3 VMs to become reachable especially when the qcow2 images are becoming bigger and bigger. This leads to spurious CI failures. Increase the timeout to 3600 seconds which should be enough for VMs boot properly Change-Id: I06942952667df28341e5ce43d30619fe2e30326b Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11bifrost: test-bifrost-deployment.sh: Use the new 'openssh-server' dib elementMarkos Chandras1-3/+3
This is a follow-up for b0a9b75e7481a24676c9488f17583c13b8b8026e ("Make use of the new 'openssh-server' dib element"). Remove the openssh-server package from the default packages and use the 'openssh-server' element instead. Change-Id: Ie14cdf93769ebefcae00f35d9e1ac149a31e7711 Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11bifrost: Make use of the new 'openssh-server' dib elementMarkos Chandras2-1/+5
Make use of the new 'openssh-server' dib element to achieve a consistent behavior of ssh server across all distributions. Change-Id: Ib10066a244fe44530969702cd465cdd79454b8fe Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11bifrost: test-bifrost-deployment: Allow overriding VM hardware optionsMarkos Chandras1-3/+3
Make VM disk, cpu and memory settings configurable to allow less powerful hosts to host the bifrost VMs. Change-Id: If819451536dd518690f3808c55de940169b90346 Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11Start using centos7 instead of centos-minimalYolanda Robla Mota1-1/+1
Current builds with centos-minimal are failing on 7, so start using centos7 element instead. Change-Id: I80d986fd7613e1807a119447a01c62e6f8a93ddf Signed-Off-By: Yolanda Robla <yroblamo@redhat.com>
2017-08-11Add step to customize create bridge templateYolanda Robla Mota1-5/+8
Change-Id: I0f4e7353773291e810f70059e4a015ac641bd2c8 Signed-Off-By: Yolanda Robla <yroblamo@redhat.com>
2017-08-11playbooks: Do not serialize ironic deploymentMarkos Chandras1-4/+4
[This is similar to upstream https://review.openstack.org/389079] This somewhat reverts 0a82e3d8d29f85229e1ed29d90de47346bc1d9b1 ("Serialize localhost actions during final host deployment"). The only point where we want to serialize the steps is when accessing common files on the host. In the bifrost-deploy-nodes-dynamic role, all tasks are accessing files based on the 'inventory_hostname' variable, which means there is no conflict there and tasks on these role can run in parallel. This allows bifrost to provision machines in parallel during pxe boot which speeds up the entire operation. Change-Id: I64b1f2832c333c89e33e8f9d19f03e4d653d74b4 Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11prototypes: bifrost: Ignore dnsmasq service failuresMarkos Chandras1-1/+1
If this is the first time we run bifrost on a host or if we haven't run bifrost since the last reboot, restarting the dnsmasq service will fail because of conflicts with the dnsmasq process running by libvirt. This is later on 'fixed' by bifrost by killing the one from libvirt. Make sure that this spurious failure will not prevent the rest of the cleanup process to move on. Change-Id: Iebf37881570c27cb07cea3cecdb0fa8ac3fc511b Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11bifrost: Serialize localhost actions during final host deploymentMarkos Chandras1-3/+8
Several tasks in the final playbooks run locally and there is a chance to encounter race conditions when multiple tasks run in parallel and all access or modify the same local files. We fix this by splitting the final steps into multiple plays and serializing those that run locally. Change-Id: I38c3465d9c4714d9db772f9ff9e32ec1d4a91db8 Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11Set the wait_timeout variable to the deployment scriptMarkos Chandras2-5/+3
There is no point in us keeping our own copy of the upstream bifrost-prepare-for-test-dynamic playbook just for overriding the wait_timeout variable. Delete the duplicate role and move the variable definition to the deployment script instead. Change-Id: Ib8066fefe80e9a3190d878eb5dba7b8d86a85e07 Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11bifrost: Refactor bifrost jobsFatih Degirmenci1-2/+3
This change - Creates job templates to verify opnfv/bifrost patches - Introduces type in order to prepare verification on baremetal - Fixes git cloning so the job shows changes for the corresponding project - Gets rid of unnecessary slave parameters from slave-params.yml - Fixes build blocking - Updates the bifrost-verify.sh script and includes in shell so we don't have to clone releng repo separately - Updates destroy-env.sh to create /httpboot and /tftpboot with permissions 755 - Removes daily infra-bifrost job and scripts as we don't use them at the moment Change-Id: Icb5a543628d3321bfe4e549ab92c67443ac578d8 Signed-off-by: Fatih Degirmenci <fatih.degirmenci@ericsson.com>
2017-08-11bifrost: Get rid of httpboot and tftpboot directoriesFatih Degirmenci1-2/+2
Change-Id: If2af5177e8614651664c7de9cde3e681ec2bb43c Signed-off-by: Fatih Degirmenci <fatih.degirmenci@ericsson.com>
2017-08-11prototypes: bifrost: README: Run the deployment script using sudoMarkos Chandras1-1/+1
It's necessary to run the deployment script using 'sudo' so that exported variables within the script are preserved in playbooks once we gain elevated privileges. Change-Id: I824b44199d57b6acf14804c3ab1825d7aa80da4e Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11prototypes: bifrost: Check we are running the destroy-env.sh as rootMarkos Chandras2-6/+14
Nearly all operations in the destroy-env.sh script need to be executed as root so make sure that's the case before we try to do anything. This also fixes the dnsmasq.leases path and we check that the file exists before we try to wipe it. The dnsmasq is also restarted to take the empty leases file into consideration. Finally, we improve the documentation to run the necessary commands using sudo. Change-Id: I56df821ca8a93307117a94a11abd0867b27d8443 Signed-off-by: Markos Chandras <mchandras@suse.de>
2017-08-11Parameterize infracloud playbook by platformsYolanda Robla Mota2-2/+22
This will allow to run on different distros depending on settings exported. Change-Id: I068d7bb2eebfd009dceeff057c8a65361483788b Signed-off-by: Yolanda Robla <yroblamo@redhat.com>
2017-08-11Update documentation about disabling selinuxYolanda Robla Mota1-4/+6
Change-Id: Ic3729d9450edabd809eed53525632e2b36752401 Signed-Off-By: Yolanda Robla <yroblamo@redhat.com>
2017-08-11Fix script when CLEAN_DIB_IMAGES var is not setYolanda Robla Mota1-0/+2
It relies on this var being exported, and if the script is ran manually and this var is not set, it crashes. Adding a default to this var to avoid this failure. Change-Id: I5c61a1d9f1aafe2d6c64af70d3237b6c5b0d1e1d Signed-Off-By: Yolanda Robla <yroblamo@redhat.com>
2017-08-11Reflect upstream change to local script fixing playbook syntax checkFatih Degirmenci1-11/+11
Upstream change: https://review.openstack.org/#/c/358477/ Change-Id: Ibfef32aec4b62d8d1d13728abec374a2d4560d81 Signed-off-by: Fatih Degirmenci <fatih.degirmenci@ericsson.com>
2017-08-11infra: Fix small doc/script issues and enable mail notificationFatih Degirmenci2-4/+8
This change - Fixes readme of puppet-infracloud - renames destroy_env.sh script - enables mail notification - disables removal of DIB images by default Change-Id: I628b26b976e38772cbbc69b737a545fe801639ba Signed-off-by: Fatih Degirmenci <fatih.degirmenci@ericsson.com>
2017-08-11Create daily job for infra deployment, remove sandbox jobs and fix bugsFatih Degirmenci2-2/+2
This change - creates upstream daily job to trigger and control the execution of jobs that provision nodes using bifrost, deployment using puppet-infracloud, and simple smoke test job. - fixes the hashbang placement in 2 of the scripts. - removes sandbox dummy jobs as the infra jobs can serve as examples now. Only the node provisioning job is functional at the moment. Change-Id: I311d159956d8b7ce3f7a500fac5a7dcb459104cf Signed-off-by: Fatih Degirmenci <fatih.degirmenci@ericsson.com>
2017-08-11Add missing curl package for infracloudYolanda Robla Mota1-1/+1
It is needed on rabbit install Change-Id: I0ebf66883cc7c66771b3b5b4c2cccc590e61e5ca Signed-off-by: Yolanda Robla <yroblamo@redhat.com>
2017-08-11Increase the disk in the infracloud vmsYolanda Robla Mota1-0/+1
The 10g disk space is not enough to spin up some vms and test. With default flavors, just a m1.small needs 20g of disk. Increasing to 100g of disk in each vm, so we can test with better conditions. Change-Id: I7b273dc5881cb6af67b7852c1318ea739217602f Signed-off-by: Yolanda Robla <yroblamo@redhat.com>
2017-08-11Increase the number of vcpus in testingYolanda Robla Mota1-0/+1
It defaults to 1 cpu, and that means that computes won't be able to spin up any node because the number of cpus are exhausted. Currently the playbook doesn't allow to individually set cpus, so defaulting everything to 4 temporarily. Change-Id: Ic6058b28f5d75a270e3897157ba034f804af8ab0 Signed-Off-By: Yolanda Robla <yroblamo@redhat.com>
2017-08-11Adjust bifrost node provisioning timeout and fix file permissionsFatih Degirmenci4-0/+4
This change - increases the node provisioning timeout to 1900 seconds since the original timeout used by upstream bifrost (900 seconds) is not suited for provisioning VMs with Ubuntu. - fixes permissions of the scripts. - removes .gitkeep. This change is not upstreamed since bifrost uses cirros for test & development purposes and it is probably not in their interest to have increased timeout. Change-Id: I4b59a321fa26f8b5895ec577a19d10f5cdcace0f Signed-off-by: Fatih Degirmenci <fatih.degirmenci@ericsson.com>
2017-08-11Add initial playbooks and scripts for bifrost POCSYolanda Robla Mota4-0/+271
Change-Id: Ic7e33f53d0111b0e833d7547fff16458506b063b Signed-off-by: Yolanda Robla <yroblamo@redhat.com>
2017-08-11Add directories for openstack-infra prototypingFatih Degirmenci1-0/+0
Change-Id: I890d36d0db54efcbcb9911707b72bf58416550b3 Signed-off-by: Fatih Degirmenci <fatih.degirmenci@ericsson.com>