fix: check lspci and dmidecode specs on x86_64 #257
Conversation
zhangqianqian
force-pushed
the
only_check_lspci_and_dmidecode_on_x86_64
branch
from
c8c1941
to
c8f4c66
Compare
| common_specs.extend( | ||
| [ | ||
| "insights.specs.Specs.dmidecode.json", | ||
| "insights.specs.Specs.lspci.json", |
There was a problem hiding this comment.
This is not correct: PCI is available on basically every mainstream and new architecture nowadays.
This is a problem in insights-core, which assumes an empty output of lspci is a bug.
There was a problem hiding this comment.
As talked in Slack, this ContentException('Empty content') is expected for insights-core. Although the empty output of lspci is valid for s390x, it (the empty lspci file in insights-archive) is useless for insights, even if it's empty on x86_64, such a ContentException will also be recorded in its meta_data JSON errors section. The errors section in the JSON file indicates why this file is not collected in the archive, but doesn't mean this is an error of the collection.
On s390x or other archs, in most cases (particular on the test VM machine), its output is always empty. so, here in this test, we can ignore check it.
There was a problem hiding this comment.
This issue is pending for a long time without update from @ptoscano . I don't see any problem showing ContentException in the log as this exception is defined for empty content on purpose. It is working exactly as what we expect in the design. Please kindly merge this request in case it blocks IQE.
There was a problem hiding this comment.
It is working exactly as what we expect in the design.
As I said in other conversations (since this was discussed outside of this PR, sigh): I consider that design broken, since it has expectations that are not actually what happens in real life systems. If the spec collection considers something that can actually happen (no PCI devices) as "error", then the collection ought to be fixed.
Although the empty output of lspci is valid for s390x, it (the empty lspci file in insights-archive) is useless for insights, even if it's empty on x86_64, such a
ContentExceptionwill also be recorded in its meta_data JSON errors section.
Then there are two ways to fix this:
- consider an empty
lspcioutput an error ONLY on x86_64; on other architectures, do not collect the empty file and do not report any error - simply send the empty file in case it is there, and let Insights/Advisory/whatever decide how to consider the situation (valid, invalid) based on the rest of the collected details of a system
There was a problem hiding this comment.
@ptoscano I think you mixed the use of "error" with "exception" in your expressions. "Exception" does not equal to "error". To insights-core, ContentException is designed for such cases like lspci and dmidcode. I don't think it is worth nor a good idea to force all Insights services to change their code logic using data collection with what the ContentException is designed for.
zhangqianqian
force-pushed
the
only_check_lspci_and_dmidecode_on_x86_64
branch
from
c8f4c66
to
21685ff
Compare
No description provided.