d64320bff6
Change-Id: I879057e401e3a15f41d9a5dfef8b65ad21e804fe |
||
---|---|---|
.github/workflows | ||
build | ||
output | ||
playbooks | ||
samples | ||
src | ||
zuul.d | ||
.flake8 | ||
.gitignore | ||
.gitreview | ||
.pre-commit-config.yaml | ||
.yamllint.yaml | ||
ansible.cfg | ||
bindep.txt | ||
errors.txt | ||
hosts | ||
LICENSE | ||
queries.yml | ||
README.md | ||
requirements.in | ||
requirements.txt | ||
requirements.yml | ||
sova-patterns.yml | ||
tox.ini |
How to get this going quickly...
- Add the query to queries.yml
- Add an associated local sova query to sova-pattens.yml
- Add a blob of text from the error you would like to test against to errors.txt
Note: the softlinks errors.txt ---> samples/errors-testing.err queries.yml ---> src/data/queries.yml sova-patterns.yml ---> src/data/sova-patterns.yml
- Execute tox
- Submit patch
- GREAT SUCCESS!
queries
Hosts reusable log queries which are built into a single queries.json file.
Query database structure
Queries are defined using the data model from src/model.py which builds a JSON Validation schema, making easy to validate the file.
One example of file can be seen at queries-example.yml
Both elastic-search and artcl can make use of regex
searches.
Pattern is supposed to be an exact string match and if multiple are present
we could easily convert them into a regex or logstash expression that uses
logical AND
.
Pattern
On elastic-rechheck queries we have cases with multiple entries used on
patterns, like message:foo AND message:bar
. This is why we also allow
a list of strings.
Categories
A query can have only one category out of a determined list of possible
values, currently infra
and code
are allowed. These can be used to
list found matches in section, making them easier to read.
Tags
Tags are also used to build the logstash queries. List of known values already used inside elastic-recheck queries:
tags:
- console
- console.html
- devstack-gate-setup-host.txt
- grenade.sh.txt
- job-output.txt
- screen-c-api.txt
- screen-c-bak.txt
- screen-n-cpu.txt
- screen-n-sch.txt
- screen-q-agt.txt
- syslog.txt
When logstash query is build OR
is used between multiple tags.
Uncovered cases:
We do not currently support the exclusions like below (2/93 found):
query: >-
message:"RESULT_TIMED_OUT: [untrusted : git.openstack.org/openstack/tempest/playbooks/devstack-tempest.yaml@master]" AND
tags:"console" AND NOT
(build_name:"tempest-all" OR
build_name:"tempest-slow" OR
build_name:"tempest-slow-py3")
query2: >-
(message: "FAILED with status: 137" OR
message: "FAILED with status: 143" OR
message: "RUN END RESULT_TIMED_OUT") AND
NOT message:"POST-RUN END RESULT_TIMED_OUT" AND
tags: "console"
To allow us to cover for corner cases not covered byt the generic format,
we could have an optional logstash
key that mentions the query. When this
would be present, we woudl avoid building the logstash query ourselves and
just use it.
Disable queries per backend
To avoid using a particular query on a particular backend we can make use of
skip: ['er', 'artcl']
.
Parameters in a query
Key | Required | Type | Sova/ER | Description |
---|---|---|---|---|
id | Required | string, unique | Sova | This id will appear in sova.log when this bug is encountered. |
pattern | Optional | string / list of strings | ER (,Sova) | This used as it is by Elastic recheck to form logstash queries like message:foo or message:foo AND message:bar If not present Elastic Recheck skips this entry |
regex | Optional | string / list of strings | Sova | Sova searches for this regex in the log files. If not present the pattern string is converted to regex (escaping special characters) and used by sova. |
tags | Optional | string / list of strings | ER | Elastic recheck uses this to form logstash queries like AND (tags:"console.html" OR tags:"job-output.txt") |
url | Optional | string | ER | Launchpad bug URL. If present the bug number and details are displayed on ER dashboard. If not the graph is titled Unknow/private bug |
suppress-graph | Optional | bool | ER | Decides whether or not to display a graph for this bug in Elastic recheck dashboard |
Adding a new item in queries.yml
To add a new query (optional: from a bug)
- Add a query item in src/data/queries.yml
- id : unique ID
- pattern: The error message(s) to look for.
- regex: If it is a complicated error message you can add a regex for sova to use.
- tags: tag(s) for logstash query
- url: Launchpad Bug URL
- suppress-graph: To display/suppress graph for the bug in ER
To move an existing sova regex from sova-patterns.json to queries.yml
-
Choose an item from
regexes
in output/sova-patterns.json. This is the file that sova currently uses. (Do not look at patterns for now) -
If the regex feels still relevant, vote for it here https://docs.google.com/spreadsheets/d/16rqgaSSoQrYNjsI4q0YInJxrOOJL3xyP3t3D6ZpluFY/edit#gid=846893892
-
Add a query entry in src/data/queries.yml
- id:
name
of the regexes item you chose. - pattern: If you want this error to be shown in ER dashboard add a string here corresponding to the regex
- regex:
regex
of the regexes item you chose.
- id:
-
Add a sample string matching the pattern (or regex if you have a special regex) in samples/errors-testing.err. As the part of check job a task calls sova to check this file and makes sure there is a match for all the regexes in queries.yml.
-
Run tox. If it passes it should create files in output/elastic-recheck and update sova-pattern-generated.
-
Commit all these files.