Age | Commit message (Collapse) | Author | Files | Lines |
|
In order to enable DPDK for OVS scenarios in Armband, we need DPDK
specific configuration to be parameterizable.
The default DPDK configuration values will remain in Fuel repo and
will be overriden with the values defined in IDF (only if present).
Since hugepage config and CPU pinning also apply to non-DPDK
scenarios, handle them together for 'common', respectively 'dpdk'.
To keep things simple, we will reuse the format expected by Fuel's
reclass model, so we can pass the whole config block as-is.
- IDF schema: fuel: Add strict checking of new block's structure,
while allowing the rvalues to be more or less freestyle;
- arm-pod6 IDF: Add initial config block implementation;
JIRA: ARMBAND-342
Change-Id: I9224f06bb6b9fb2387414eb2a1be35461c27fb99
Signed-off-by: Cristina Pauna <cristina.pauna@enea.com>
|
|
While at it, order installers by name in IDF schema.
Change-Id: I4bf1dd047a3f7a64b8bc928951fc18c4728a9b68
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
MaaS timeouts should be configurable on per-POD basis, since nodes
in some labs take longer to comission/deploy than our default values.
IDF SCHEMA:
- add new properties to Fuel IDF section;
Change-Id: Ife5ff85faeae683a8c1781cae8b29f09c1f07b82
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
There is no easy way to determine automatically if a PDF/IDF set is
compatible with a specific installer adapter.
Most common incompatibilities are caused by:
- installer adapters hardcode more network ports than nodes have
defined in PDF;
- installer adapters hardcode more nodes than PDF defines;
Previous attempt at handling this dynamically (cd54edf) fails for
recently added PODs, so let's handle this explicitly per-pod via a
new optional IDF property, 'idf.installer'.
If 'idf.installer' is defined, it will be used by `check-jinja2.sh`
as a filter-in list of installer adapters to check against.
The default (no 'idf.installer' present) is to check all IAs.
Leverage the new property for:
- arm-pod5: not enough network ports on computes for Compass, Joid;
- arm-pod7: same as arm-pod5;
- bii-pod1: no 'storage' network defined in IDF for Daisy;
- zte-pod9: 3 cluster nodes are only supported by Daisy for now;
SPEC changes:
- net_config (previously mandatory for IDF version '0.1') is now
mandatory only if 'idf.fuel' or 'idf.daisy' is defined;
Fixes: cd54edf
Change-Id: I706c05519e5491ad631069d7cc090e9c8bfd3011
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
- add basic structure for IDF schema validation;
- fill in Fuel section;
- add stub for Daisy section;
- add stub for Compass section;
- add stub for net_config;
- enable IDF schema validation via `check-schema.sh`;
- prepare for new IDF versions with different schemas,
while allowing subsection schema reuse from IDF v0.1;
- extend validate_schema.py for IDF;
Change-Id: I59b1c9d857e9dee9e6ff9956e81b2ae6dc27a767
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
- add 'version: 1.0' to PDF spec;
- add 'version: 1.0' to all existing lab PDFs;
- extend schema with new property;
- add workaround for value-based decision-making in schema version
selection via `validate-template.py`;
- add support for multiple schema versions;
- add versions for all schema blocks defined so far;
- fix PDF schema pattern for disk size decimals (e.g. '1.8T');
Change-Id: Ie8f768803ec19f1f9a7982fe5ca59df80764fc4a
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
|
|
zte-pod3 and ericsson-pod1 define their disk interfaces as (i)SCSI.
Change-Id: I97945f7a41eec4e8a38de544af95ce4af82df200
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Jumpserver and cluster nodes use different keys for defining the NIC
name ('nic' vs 'name'), switch all of them to 'name' for uniformity.
Change-Id: I2d7720f5e5349ef59cf76a0e07749cfbd0d34d0b
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
ssd and nvme drives are static, so extend the list of accepted
disk_rotation values to accomodate this.
Use '0' so the enum can still be parsed as a numeric value.
Change-Id: Ib8ef8cceb495c7eb588c68d151690747463167f5
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
- create new YAML schema for PDF validation;
- add basic python script for checking a PDF against the schema;
- add bash wrapper for checking all PDFs in Pharos, to be leveraged
later via a new verify CI job;
Change-Id: I47e02642756b7a231138dec3d5258b100b4db72b
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
This only covers the `config` subdir.
Change-Id: I1fb9470e216cc2de9a4ac91de44d99e7044e708e
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Fix typo before it gets propagated even more:
$ grep hasewell -Rl . | \
xargs sed -i 's/hasewell/haswell/g'
Change-Id: Id6a17ae8f429946f7febba6f319a25759fc8339f
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Note: IDF data encryption is not supported. Supporting that is
trivial, but it leads to slightly more complicated code, plus it
breaks support for multiline scalar encrypted data in the PDF ('>'),
forcing us to define each encrypted value as inline string.
While at it, fix silly limitation of jinja2 path residing in a subdir
of CWD.
Change-Id: I441ec754d8b6e4aad2ed73aba0b9b18ed65f05f4
Signed-off-by: agardner <agardner@linuxfoundation.org>
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: Ibb481d57e3c154c409c2635096330f2e63dce9db
Signed-off-by: Jack Morgan <jack.morgan@intel.com>
|