summaryrefslogtreecommitdiffstats
path: root/mcp/reclass/classes
AgeCommit message (Collapse)AuthorFilesLines
2018-07-02[apt] Allow downgrades during APT operationsAlexandru Avadanii2-0/+4
Pinning certain packages (e.g. OVS) interfers with `dist-upgrade` if downgrades are not explicitly allowed. Change-Id: I0b001936e86228037c30f30812e32d6221ef458d Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-07-02[AArch64] Pin armband repo with highest prioAlexandru Avadanii1-0/+8
Mirantis repositories shadow certain packages from Armband repos (e.g. qemu-efi) if pinning is not set. Since MaaS does not allow configuring repo priorities via its API, use Salt pillar data to re-define the Armband repo and set its prio. Change-Id: Ic743f4b684b3f13552792f5f04097fac73171b37 Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-06-29Revert "[salt-formulas] Switch to 'testing' dist"Alexandru Avadanii2-2/+2
This reverts commit a8f434118013a69295bdc0f9fc8294c539742c01. This reverts commit b52ef6ea62f43de4b91e2becd92a3282e569925c. Change-Id: I604df3f50febb049f1545a8141c5abb013cf6c69 Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-06-29[patch] MaaS: fabric ID from CIDRAlexandru Avadanii1-3/+0
Change-Id: I1e1490f4d0a0bee2783450c6369d7ab6c45b193f Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-06-29[salt-formulas] Also switch reclass to 'testing'Alexandru Avadanii2-2/+2
Previous commit only changed the dist during Salt master bootstrap, but failed to update the DEB repo configration in reclass. Fixes: a8f43411 Change-Id: Ie011719cdc1c75cba363f53e51f22ce7cf138a5e Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-06-26Merge "[ha] Add fabric to maas admin subnet"Alexandru Avadanii1-5/+8
2018-06-26[ha] Add fabric to maas admin subnetMichael Polenchuk1-5/+8
Change-Id: Ida4cf736dfe60886384e7e180d3e43faf811bd58 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-06-25[ha] Pin UCA's libvirt/qemu on computesMichael Polenchuk4-7/+7
* install libvirt/qemu packages only from UCA on computes * shift panko/aodh db settings to telemetry nodes Change-Id: I0f4434aa8d2139870afaed9d9831db0a9af10e51 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-06-25[MaaS] Adopt boot-resources control from maasngAlexandru Avadanii1-2/+19
JIRA: FUEL-364 Change-Id: I891514f85cf694509003a3b0a6f3568524d0a461 Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-06-22[ha] Update pymysql to version 0.8Michael Polenchuk3-10/+38
* get latest pymysql from UCA repo * set DB connection recycle time to avoid issue with broken database connection during query (MySQL server has gone away) * set haproxy source balacing mode for neutron API to avoid race condition with duplicate security groups * get libvirt/qemu from UCA repo on compute nodes since main process of libvirt exits (is killed by 11/SEGV [segfault]) during volume attaching operation Change-Id: Iead5ecc561683876c476357f4a4de8780775aa7b Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-06-21[ha] cmp: Fix duplicate PXE/admin IDs (2)Alexandru Avadanii3-0/+15
Certain scenarios build additional interface definitions which overlap with the ones from common-ha reclass dir. Extend previous fix to scenario-specific interface definitions to prevent such duplicates. Change-Id: I57b175e2f760a22454723266b044845335af4cfe Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-06-20[MaaS] curtin: Switch back to Salt 2016.11 repoAlexandru Avadanii1-1/+1
Change-Id: I7482a3b7045e6e1e94d6abfe2dc6b862f2176a53 Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-06-20[noha] Rectify opendaylight+dpdk scenarioMichael Polenchuk3-9/+10
Change-Id: Ie707de90617e7080a5eb87cad69a604cf5e0f9fc Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-06-20[ha] cmp: Fix duplicate PXE/admin interface IDsAlexandru Avadanii1-0/+6
Change-Id: I99a8acfbfdddb79493bf3a191835927cef5fd5e3 Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-06-19[queens] Update Gnocchi versionMichael Polenchuk2-2/+2
Change-Id: Ieea2c510ae4c6d4db0667f3341c877042a2c1846 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-06-19Enforce static configuration instead of DHCPAlexandru Avadanii47-519/+455
- noha: 'accept_policy: open_mode' to align with ha scenarios; - s/cmp01/cmp001/g to align all scenarios and allow code reuse; - rename network params: s/dhcp/mcpcontrol/g, cleanup; - computes XDF data: drop 'opnfv_*' layer of params, cleanup; - local vPDF: add comments with default roles by node index; - parameterize all netmasks; - drop unused address/netmask for 'proto: manual' interfaces; - virsh_net: cleanup definitions, remove hardcodes, align IP on jumpserver and DHCP range with MaaS for pxebr; - maas: parameterize hardcoded '/24' cidr for PXE/admin, refactor maas.region.machines parameterization; - merge <all-mcp-arch-common/infra/config_*pdf.yaml.j2> templates; - move reclass.storage definitions of compute nodes to common dir; - drop 'openstack_compute_*' reclass params in favor of expanding them via j2 directly in reclass.storage params; - adopt `nm.cluster.has_*_nodes` where possible; - obsolete `runtime.yml` from reclass model; - refactor arch-specific reclass param selection; - remove unused defaults in favor of mandatory IDF properties; - noha: prepare for baremetal node support in cinder_lvm_devices; - interfaces: add interface_mtu and 'noifupdown: true' everywhere; - interfaces: use j2 macros to generate eth/vlan config; - states cleanup: remove DHCP route disable workaround on prx/cmp; - allow configuring NTP servers via: `idf.fuel.network.ntp_strata_host{1,2}`; - ovs_bridge: Allow setting gateway, dns-nameservers - apache: Adjust module list for novcp class inheritance; - glusterfs PPA: pin with same prio of MCP repos for novcp scenario; JIRA: FUEL-319 JIRA: FUEL-326 JIRA: FUEL-337 Change-Id: Ia6ad64ba8cade85a75fb22c9a2505decc3834360 Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-06-18Rectify nova placement API checkMichael Polenchuk5-16/+5
* make a GET request to placement service as it returns 200 with version data starting from nova 17.0 * turn on nova image preallocation mode on baremetal computes * double salt worker threads by number of available CPU Change-Id: I509cbab3b4b8460452f5e487e457edfedd5da29a Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-06-14Get back to salt 2016.11Michael Polenchuk4-2/+6
Change-Id: Id024ed22dd1760f41ae18aeb8e680c2f07a5dc63 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-06-11[ha] Fix salt state apply on ctl VCP nodesAlexandru Avadanii1-1/+1
Sphinx metadata collection does not check whether a haproxy server is enabled or not, so keep full 'heat_cloudwatch_api' configuration in place to prevent breakage; and just disable it instead. Change-Id: I9fda1f3d2829b89ea8e425ae831a54484651f60a Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-06-11Smooth down API proxy configMichael Polenchuk3-14/+17
* heat: remove deprecated AWS compatible CloudWatch API * nova: placement API responds with 405 when doing GET to root URL Change-Id: I545e5b0fc5b57b516a4b1d8c9e33ea1d5d61ed76 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-06-07[ha] Update OpenStack version to QueensMichael Polenchuk97-333/+353
* make cluster names version-agnostic * switch to mirror.mirantis.com * system reclass update Change-Id: I96394b3c9d8d38321289af8e3a4e8bbc5dd7a408 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-06-07[noha] Update OpenStack version to QueensMichael Polenchuk44-65/+81
* make cluster names version-agnostic * pre-install oslo-templates formula * salt version update to 2017.7 * system reclass update Change-Id: I903895f6bcd98a1062e66e3f14d23ec04dc1b869 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-06-07[patch] Rebase APT keyserver proxy support patchAlexandru Avadanii2-2/+4
Sync our patch with upstream PR [1]. [1] https://github.com/salt-formulas/salt-formula-linux/pull/138 Change-Id: If3a51d3aef51accee0901c59e1342b42dbe4be4b Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-05-31[noha] Disable proxy certificate generationAlexandru Avadanii1-1/+0
noha scenarios try to generate the proxy SSL cert using the Salt Master CA authority. However, this requires a `salt-minion` restart between the CA configuration and the cert generation (see [1]). Since we don't use this cert anyway, let's just disable it for cfg01. JIRA: FUEL-377 [1] https://github.com/salt-formulas/salt-formula-salt/issues/67 Change-Id: I8e88200e0d5614f0034c0c77e6ba675dc7ef0c11 Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-05-24Merge "AArch64: Switch back to hwe-16.04 kernel"Alexandru Avadanii1-0/+8
2018-05-24Merge "[maas] Adopt maas.machines.storage"Alexandru Avadanii1-0/+8
2018-05-24maas: boot-resources: Dynamic arch list from PDFAlexandru Avadanii1-1/+3
Instead of leaving MaaS default to 'amd64', explicitly set the list of architectures to support based on PDF data. Change-Id: I852a3ce156db3df0c090f10c0b45c26058dbb6c6 Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-05-24[maas] Adopt maas.machines.storageting wu1-0/+8
Replace MAAS CLI set_disk_layout with the new maas.machines.storage state JIRA: FUEL-364 Change-Id: I4d8cd9f473c5386ee7b32ad378ca1e02989233ca Signed-off-by: ting wu <ting.wu@enea.com>
2018-05-23AArch64: Switch back to hwe-16.04 kernelAlexandru Avadanii1-0/+8
Align kernel versions across architectures (where possible). Change-Id: I66a822611eb5e46f90b62f5d36df571ae75dcba3 Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-05-22[maas] AArch64: Preseed Armband repositoriesAlexandru Avadanii2-14/+52
Preseed Armband common repo-comp for Pike, so we get the updated kernel & other packages from the start. Requires upstream PR [1]. While at it, also handle related PR merged upstream [2] by explicitly setting maas.region.subnets name via j2 instead of reclass param expansion in name (allows us to drop the remaining chunk of the fabric-from-CIDR patch we used to carry). Note: opnfv_infra_maas_pxe_network_addres can now be dropped from pod_config j2 template in Pharos Fuel installer adapter. [1] https://github.com/salt-formulas/salt-formula-maas/pull/26 [2] https://github.com/salt-formulas/salt-formula-maas/pull/22 Change-Id: I356adb73b80f0f2d85db4ab060e804cb9a053862 Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-05-21[ha] Bring in gnocchi/panko servicesMichael Polenchuk6-12/+84
JIRA: FUEL-372 Change-Id: I9083898f0179649d855750d48f3f9bc79caba50c Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-05-18[ha] Align package repositories on all the nodesMichael Polenchuk5-16/+3
Change-Id: Idd09d28677b109b4c0c5a4c232dcfb7c4f24dbd7 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-05-17[noha] Bring in gnocchi/panko servicesMichael Polenchuk3-10/+41
JIRA: FUEL-372 Change-Id: I4e322a4a2c84843e9350fe9b3b849cd0c5244a12 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-05-16[noha] Align package repositories on all the nodesMichael Polenchuk4-42/+7
* replace UCA repo with MCP one on the remaining nodes * clean up haproxy out-of-use frontends: - Nova EC2 API - Glare API Change-Id: If8e6de2d95387c92ae96ad6069ec788ab229e1ce Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-05-09[xdf] Fix parsing of non-range private vlan(s)Alexandru Avadanii4-3/+5
Private (tenant) vlan range support introduced a regression in parsing PDF/IDF with a single vlan (no range) due to the value being interpreted as an int instead of string. Change-Id: Id5d67c4c516a9bb807ffa5e111ead5cbcee7e2ca Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-05-08[virtual] Bring in OpenDaylight DPDK scenarioMichael Polenchuk2-1/+23
Change-Id: If77ac85fa86e0a1a18c0cc2abff77d876cdb9e93 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-04-28[odl] Remove unused connecter/port from proxyMichael Polenchuk2-23/+0
Change-Id: I7a880c0a011d6f5f4a7e2d8c8309871b1c290235 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-04-24Merge "Mend OVN scenario"Alexandru Avadanii1-0/+3
2018-04-24Update qemu/libvirt on compute nodesMichael Polenchuk2-0/+2
Get qemu package with fix of TSC frequency mismatch between instance and host which breaks up nova pause/unpause operation. JIRA: FUEL-368 Change-Id: I2c954b2b0a8cddca30c861f565b63b642e0b0e9f Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-04-24Mend OVN scenarioMichael Polenchuk1-0/+3
* setup HWE kernel to get suitable conntrack module * clean out outdated state with ovn ctl options * point SB remote source to local mgmt network Change-Id: I8986c227ce0a9a3b7ab3faf382760ec32e6e7c00 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-04-22[ha] Fix tenant NIC OS config for VLAN rangesAlexandru Avadanii3-6/+9
Previous commits added support for tenant VLAN range parametrization, yet failed to accomodate the new VLAN range representation ('M:N') in existing `linux.network.interfaces` reclass configuration. This lead to the silent removal of <nic_private>.<vlan_private> from interfaces file and for `br-mesh` to try to use <nic_private> instead of <nic_private>.<vlan_private>, conflicting with `br-floating`. Stick to using the first VLAN in the tenant range. JIRA: FUEL-365 Change-Id: I88f7263159f1742dba3b6e36649c9344f2932c5f Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-04-19Disable cpu pinning & L3 HA featuresMichael Polenchuk5-9/+5
* CPU pinning: tests have no support for dedicated CPU policy, so enabled pinning causes general performance impact [https://docs.openstack.org/nova/pike/admin/cpu-topologies.html] * L3 HA routers: - doesn't catch l3-agent failure, relies only on HA network state to spot failure - failover process only retains the state of network connections for instances with a floating IP address - incompatible with DVR mode JIRA: FUEL-360 Change-Id: Ie0182bf953b9989729f89d705d9fee902d229e51 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-04-18[reclass] Parameterize tenant VLAN rangesAlexandru Avadanii2-2/+2
Instead of hardcoding the VLAN ranges in reclass, read them from IDF. This is required when VLAN segmentation is used instead of VXLAN, as lab switches should be configured appropiately (e.g. Armband PODs use VLANs 2000+ instead of 1000-1030). Change-Id: I009ccea9b328c40f055a745791eafe3393b96390 Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-04-11Clean up opendaylight configurationMichael Polenchuk5-12/+0
* remove client patch (has been merged to upstream) * websocket now is disabled by default * purge deprecated/unneeded features (odl-mdsal-apidocs, odl-dlux-all) Change-Id: Id247a3c74e66730049bfeafc4a5164aad0d9b662 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-04-09Add ports to ovs bridge configMichael Polenchuk1-0/+2
To avoid race with interfaces initialization specify all the ports that belong to a bridge in the config file. Change-Id: Ie4f48e7a81d4659a0f43749cdd518dbeffe2e35f Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-04-06Merge "[virtual] Set vnc address to local one"Alexandru Avadanii1-0/+2
2018-04-06[virtual] Set vnc address to local oneMichael Polenchuk1-0/+2
System reclass model assigns vnc address to control_address variable which is used for controller node reference, so change it to local address of the current compute node. Change-Id: I0c6b8534a1dae516620d0d250944232379a526e0 Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
2018-04-06[UCA] Stop enforcing repository architectureAlexandru Avadanii2-4/+0
Let APT automatically choose the UCA/Pike-staging PPA architecture based on current system architecture instead of hardcoding it. Change-Id: I4fe645847ac0fb70a852477feba9f788fcb8da81 Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
2018-03-31Merge "[odl] Configure Odl memory requirements"Alexandru Avadanii2-3/+3
2018-03-31Merge "Add pike staging repository"Alexandru Avadanii2-0/+12