Unify base images to reduce disk image size #182
Labels
Comments
|
Guess we should benchmark this by providing size tests of images before and after this work to see if there is an appreciable difference |
This was referenced Oct 22, 2021
shawaj
added a commit
to NebraLtd/hm-diag
that referenced
this issue
Oct 23, 2021
- change base image across as many containers as possible to reduce image size - unpin versions - fix package names (due to change from alpine to Debian) - ignore pinning warning - call correct pip - remove unused installations - fix --no-cache-dir call - bump hm-pyhelper - combine some RUN commands to reduce layer size Relates to: NebraLtd/helium-miner-software#182
|
So changing UPnP from Alpine to Raspberry Pi Debian Python increased the size of the container from 47MB to 287MB - however this does not take into account the fact that this base OS image is shared across upnp, hm-diag, hm-config and soon also hm-pktfwd...
Diag has actually reduced in size from 327MB to 313MB changing from alpine to debian-python...
|
shawaj
added a commit
to NebraLtd/hm-pktfwd
that referenced
this issue
Oct 26, 2021
Bump to more up to date base image NebraLtd/helium-miner-software#182
shawaj
added a commit
that referenced
this issue
Oct 27, 2021
Bumps UPnP, config, diag and packet forwarder containers so they are now all using the same base image. We need to now test that it's actually smaller so will leave #182 open for now.
Closed
vpetersson
pushed a commit
that referenced
this issue
Nov 15, 2021
…and pktfwd rewrite (#215) * test diag with new base image test diag with new base image * bump upnp for testing bump upnp for testing * fix: add depends on for packet forwarder The packet forwarder actually requires the miner container to be able to accept connections on the hostname helium-miner before it will be able to start the concentrator. This adds a depends_on to not attempt to start the packet forwarder container until the miner container is up * Create docker-compose.yml Bumps UPnP, config, diag and packet forwarder containers so they are now all using the same base image. We need to now test that it's actually smaller so will leave #182 open for now. * Bump to 2021.10.27.0_GA Bump to 2021.10.27.0_GA * Added documentation to the readme for quick start (#189) Added documentation to the readme for quick start (#189) * Bump to 2021.10.29.0_GA Bump to 2021.10.29.0_GA * Bump to 2021.11.02.0_GA Bump to 2021.11.02.0_GA * rearrange readme rearrange readme * Update README.md * Bump diag and config Bump diag and config * feat: bump GA to 2021.11.04.2 * release: bump GA to 2021.11.10.0 * release: bump GA to 2021.11.11.0 * Add tech support to CODEOWNERS Partially fixes #208 * fix: ECC locking and initFile (#209) Lock ECC usage and ensure initFile always returns. * build: bump firmware version * bump(pktfwd): current master branch Including likely unnecessary patch to regions. Co-authored-by: Aaron Shaw <[email protected]> Co-authored-by: Sebastian Maj <[email protected]>
|
This work is now complete. Will close this issue for now 👍 |
|
For reference, this reduced the zipped disk image size from ~600 MB to ~400 MB Overall image size has gone down from 2.79 GB to 1.83 GB So a healthy saving! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
In discussion with @vpetersson we have discussed using the same base image across diag, upnp, config and pktfwd containers to reduce the overall balenaOS image size
This is a tracking issue for this work and unifies work from these other repos:
Currently the miner and dbus containers are built by helium and balena respectively so we would need to build these ourselves to use a single base image for all containers. This is probably out of scope for now but may be worth revisiting in the future.
The text was updated successfully, but these errors were encountered: