The retry logic was meant to work around issues where a round-robin
redirect of mirrors gave us random mirrors of varying quality. This was
not used in practice, rather fixed mirrors were always used (either
hard-coded as basurl, or resolved from metalink).
The retry logic meant that when we did hit very slow mirrors we would
time-out and retry, potentially failing altogether, even though the data
was coming. Each retry would not help, as the mirror was anyway the
same. As a result our CI gave us avoidable false negative test results
some of the time.
The proper solution to this is to gain support for librepo and metalinks
to adopt the same retry logic that dnf uses.
For now, improve on the retry logic by retrying until a max total time,
rather than an increasing timeout on each try. Up the given timeouts to
be one minute to connect and five minutes to complete the download. This
avoids hanging forever if the mirror is truly broken, but still gives
more time to finish the download than each iteration in the old code
did.
There are no new tests for this, as before this change the tests mostly
passed, and after it they will hopefully still mostly pass (but more
often).
Signed-off-by: Tom Gundersen <teg@jklm.no>
We used to have a try-except block to catch URL requests that are not in
`urls`. This block has since then grown way bigger than it should be. We
may now accidentally catch KeyError exceptions from lots of other
places.
This commit extracts the accessor of `urls[checksum]` and saves the
result in a local variable and makes the remainder use that variable.
When the urls' secrets field is not set, an error is thrown when trying
to get the name of the secrets. The secrets now have a default value of
{} when they are checked for the name.
When osbuild is given a manifest, the sources' urls can contain fields
for both a url path and a secret for that url. If the secret is
org.osbuild.rhsm the system's rhsm certificates are retrieved. These
certs are included when the files are curled.
Add a 10s connection timeout for each file transfer. Also add an
increasing max timeout for a given file transfer (30s to 180s).
Also increase the retries to 10 and the concurrent threads to 15.
Hopefully this should make things a bit more stable in the face of
bad mirrors. We were encountering mirrors that would hang either
on connect or download at such slow speeds that they might as well
have stalled (~1kB in 45s).
Follow-up patches will provide a more long-term solution, by
allowing the same mirror selection as dnf currently uses.
Signed-off-by: Tom Gundersen <teg@jklm.no>
This source adds support for downloaded files. The files are
indexed by their content hash, and the only option is their URL.
The main usecase for this will be downloading rpms. Allowing depsolving
to be done outside of osbuild, network access to be restricted and
downloaded rpms to be reused between runs.
Each source is now passed two additional arguments, a cache directory
and an output directory. Both are in the source's namespace, and
the source is responsible for managing them. Each directory may
contain contents from previous runs, but neither is ever guaranteed
to do so.
Downloaded contents may be saved to the cache and resued between
runs, and the requested content should be written to the output dir.
If secrets are used, the source must only ever write contents to
the output that corresponds to the available secrets (rather than
contents from the cache from previous runs).
Each stage is passed an additional argument, a sources directory.
The directory is read-only, and contains a subdirectory named after
each used source, which will contain the requseted contents when
the `Get()` call returns (if the source uses this functionality).
Based on a patch by Lars Karlitski.
Signed-off-by: Tom Gundersen <teg@jklm.no>
