toggle quoted messageShow quoted text
From: cncf-toc@... <cncf-toc@...> on behalf of Brian Grant via lists.cncf.io <briangrant=google.com@...>
Sent: Wednesday, October 14, 2020 7:05:41 AM
To: Liz Rice <liz@...>
Cc: Chris Aniszczyk <caniszczyk@...>; Matt Farina <matt@...>; agupta <agupta@...>; Michelle Noorali <Michelle.Noorali@...>; Priyanka Sharma <psharma@...>; cncf-toc@... <cncf-toc@...>;
Saad Ali <saadali@...>
Subject: [EXTERNAL] Re: [cncf-toc] [URGENT] Impending deletion of Helm 2 releases and charts buckets
On Wed, Oct 14, 2020 at 7:02 AM Liz Rice <liz@...
Thanks Chris. Just noting that we do have the Special Issues process where $$$ are involved to take technical budget issues to the GB, though it sounds unlikely that this would be an affordable item for the community to sustain. (I am wondering
if there is a stop-gap measure to fund the storage for a short period of time while the deprecation message gets through to users, to stop it falling off a cliff - but there is only any point in doing that if we can effectively encourage migration off of these
This has been an issue since at least April 2019. The deprecation message is not getting through.
On Wed, Oct 14, 2020 at 2:20 PM Chris Aniszczyk <caniszczyk@...
It would be great to have input from the current Helm maintainers on their issue and public deprecation plan here. +Matt
Also, this is a big budget item and will likely involve the GB, where the TOC should mostly focus on the technical deprecation plan and how to do this for the overall ecosystem with minimal impact. As I understand it, the storage cost of the charts is
relatively cheap but the bandwidth costs are considerable and would make up a huge chunk of the budget and honestly something usually sponsored by large cloud providers or a CDN, similar to how we handle the Kubernetes project or even other projects like say
A simple solution would be to reuse or extend the existing credits that exist for Kubernetes  and firm up an ownership plan with the Helm maintainers on when deprecation can happen with a firm date, but this requires collaboration with the Helm maintainers
 and the wider community . We can also try to offer some time at the upcoming KubeCon to further make the community aware of the upcoming deprecation, but these things are always tricky for when it comes to widely used software.
On Wed, Oct 14, 2020 at 12:51 AM Liz Rice <liz@...
It does seem like a hard cut-off is going to hurt people without some further action. I can see two ways the CNCF might be able to help:
- Getting the word out. I suspect that a lot of Helm chart creation is done by folks outside of the "Helm community" and maybe the message is not reaching the wider cloud world. This isn't a criticism of the Helm community or the way it communicates,
it's an indicator of how widely used the project is. I think CNCF could help push a message to this wider audience
The affected charts are mainly the ones in the official community repo:
There is a notice in the README.
PRs to the charts are still being merged, though at a lower rate than earlier in the year:
- I'm not clear from the GH thread whether there is already a CNCF owned location that the old locations can be redirected to, but I'd be supportive of CNCF funds being used for that purpose. Is it more-or-less as simple as redirecting URLs
from google-funded resources to community-funded resources?
There is not a CNCF-owned location that I'm aware of.
"The URL to the stable repository has been embedded in Helm since v2.0.0-alpha.2. It's directly to a Google Cloud bucket"
No redirector was put in place.
The project containing the buckets could be moved to a CNCF account.
Does Artifact Hub have a role to play here?
Finally, I can't be the first person to think this thought, but is there some possible way to contact chart owners and tell them about the imminent demise of the current buckets?
On Wed, 14 Oct 2020 at 07:45, Arvind Gupta <agupta@...
I would recommend setting up a URL forwarding from old repos to new repos, so that charts are maintained at new locations but old URL references continue to work.
This has the potential for causing a lot of pain to users.
The buckets are scheduled to be deleted on Nov 13, 2020, per the
original deprecation plan
. As that date is fast approaching, it is becoming clear that doing so could cause major issues. Per
on the bug "The usage of the charts storage buckets hasn't decreased at all. It's higher than when the deprecation was announced last October. There have been about 3.5 accesses per second
on average over the last 12 hours."
But the hard part to fight is the countless posts/examples on stackoverflow/other sites, and
the many vendors who apparently haven't migrated yet.
Given that and the bucket usage metrics, the fear is that the number of people still using the old repos may be under-estimated, and shutting down the buckets could cause many unanticipated outages. The long
lead time was supposed to be about getting users off, but that result does not appear to be achieved yet.
As TOC, should we consider alternative options such as transitioning ownership of these resources to CNCF to ensure a longer runway for deprecation?