timescaledb-docker-ha
timescaledb-docker-ha copied to clipboard
Bug Fix: Missing Barman Cloud entrypoints
In #442 I added the python3-barman
package in anticipation for the ImageCatalog
feature in CloudNativePG that would allow running the official Timescale image in CNPG.
That PR though has two issues:
- Unlike
python-pip
,apt-get
does not install package binaries, required for interaction withbarman-cloud
. - It did not add necessary dependencies for Azure and Google cloud support.
I tried implementing it without python-pip
, but the Azure Storage package python3-azure-storage
is over 600MB
which completely defeats the purpose of the size optimization. With pip
the barman
package only fetches azure-blob-storage
and this results in around 300MB
smaller image, compared to the deb package only approach.
I suspect that once I add python3-pip
it may be beneficial for other as well to switch to pip
packages instead of Debian packages as:
- More targeted packages have a smaller size impact and less bloat
- Newer versions of packages are available via
pip
.
In fact if you request it, I will move the apt-get install -y python3-pip
to the beginning of the Dockerfile so it is available to use earlier. I only kept it here so the Docker image is easier to build and use more cache.
P.S. Special thanks to @NiccoloFei for his assistance in debugging this.
@graveland Can I trouble you with a review, again?
@feikesteenbergen @graveland I would like to release support for TimescaleDB in the official CloudNativePG chart next week. We have everything else prepared, but this PR.
Would you be able to take allocate some time for a review?
@graveland Would you like me to make that change now?
@graveland Would you like me to make that change now?
If you don't mind, thanks!
@graveland I scanned through the Dockerfile and this seems as a good spot. I also explicitly added python3
so it's not marked as auto, as it is required at runtime.
@graveland
P.S. I can't marge this, but ping me if you would like me to make any other changes.
All good, thanks!