Hi folks, It was great meeting some of you in person at Community over Code where we had a chance to discuss the Cassandra Sidecar project. A lot of you expressed interest in having a release of Sidecar to start using the existing features in the project:
- C* adapters for versions 4.0, 4.1, 5.0 and trunk - Cassandra Analytics support - Restoring SSTables from S3-compatible object storage - Mutual TLS authentication - Role base access control - Cluster health checks - Observability - Sidecar Client Some of the call-outs in this thread of things we need for a release: - Documentation - Bug fixes [1][2][3] These are other things mentioned in the thread, where we would like help from the community: - vNode support [4] - Backup support [5] - Bulk APIs [6] Once we have documentation and the bug fixes mentioned above, I will start a thread vote for a release of Cassandra Sidecar 0.1 alpha. We want the community to benefit from the features already present in Sidecar. The more community engagement we have, the more feature-rich will become. Additionally, we can leverage easy-cass-stress to have Cassandra Sidecar included in your Cassandra clusters, so we can easily start trying it out. Please let me know your thoughts on this. Best, - Francisco [1] https://issues.apache.org/jira/browse/CASSANDRASC-120 [2] https://issues.apache.org/jira/browse/CASSANDRASC-121 [3] https://issues.apache.org/jira/browse/CASSANDRASC-122 [4] https://issues.apache.org/jira/browse/CASSANDRASC-149 [5] https://issues.apache.org/jira/browse/CASSANDRASC-148 [6] https://issues.apache.org/jira/browse/CASSANDRASC-3 On 2024/10/03 12:05:14 Josh McKenzie wrote: > I'm tentatively in favor of the idea of us cutting releases for all our > ecosystem dependencies as a blocker to cutting a major with Cassandra proper. > I say tentatively since we've had trouble getting majors across the line for > awhile so adding more dependencies to that feels risky, however I think the > improvement to user experience justifies it. > > Also - I suspect the effort required to get subprojects across the line will > be quite a bit less than the mainline DB. > > If this is something there's agreement / consensus on, I'd be happy to take > on the role of driving that coordination in the future. > > On Wed, Oct 2, 2024, at 3:16 PM, Jon Haddad wrote: > > > Mostly Analytics, which should not be a blocker for Sidecar. > > > > Yes, agreed. I'm just trying to understand the context of the vnode > > statement and how we're framing 1.0 sidecar. > > > > > We definitely need to support 5.0 for the Analytics release, but that's > > > orthogonal to Sidecar. > > > > It is, unless we're endorsing the analytics library as a primary reason to > > use sidecar. Then it becomes a dependency people rely on, and I don't want > > it blocking people from upgrading. > > > > > Yeah, definitely. I agree that we need to support the latest released > > > version of Cassandra in ecosystem projects. However, without official > > > releases there won't be adoption and without adoption there won't be > > > feedback from the community on what features / improvements are needed. > > > > I don't expect that our first release will be feature complete, but it > > should be at least compelling. I'm still not aware of what functionality > > exists that would meet that requirement. > > > > Think about this from the perspective of reading a blog post. We're > > excited to announce sidecar 1.0! Here's what you can do: > > > > 1. Backup / restore? > > 2. ? > > 3. ? > > > > All I'm asking for are 3 reasons why people should care. If one of them is > > backups, do we provide more flexible backup targets than S3, and if we > > provide incremental / continuous backup options? Is there a scheduler or > > do people need to roll their own? Is it coordinated, or is the intent that > > people handle it on their own? > > > > I work with a lot of end users - this is the type of thing that affects > > people and can either help or harm the project image. > > > > Jon > > > > > > On Wed, Oct 2, 2024 at 2:40 PM Francisco Guerrero <fran...@apache.org> > > wrote: > >> > By vnode support, do you mean the analytics library? Or do other > >> > features > >> > in sidecar not work with vnodes? > >> > >> Mostly Analytics, which should not be a blocker for Sidecar. However, I do > >> feel there should be more testing around vnodes in Sidecar. > >> > >> > If we're talking about analytics, that gets a little complicated. Are we > >> > also then talking about 1.0'ing analytics? If so, I think we need > >> > support > >> > for 5.0 and BTI there. > >> > >> We need to release Analytics at some point as well, we should have a > >> similar > >> discuss thread regarding Analytics. We definitely need to support 5.0 for > >> the > >> Analytics release, but that's orthogonal to Sidecar. > >> > >> > Increasing the size of the ecosystem is challenging... > >> > >> Yeah, definitely. I agree that we need to support the latest released > >> version > >> of Cassandra in ecosystem projects. However, without official releases > >> there won't be adoption and without adoption there won't be feedback from > >> the community on what features / improvements are needed. > >> > >> On 2024/10/02 18:05:42 Jon Haddad wrote: > >> > By vnode support, do you mean the analytics library? Or do other > >> > features > >> > in sidecar not work with vnodes? > >> > > >> > If we're talking about analytics, that gets a little complicated. Are we > >> > also then talking about 1.0'ing analytics? If so, I think we need > >> > support > >> > for 5.0 and BTI there. > >> > > >> > In my opinion, if something core to the project has official releases, > >> > such > >> > as drivers or operational tooling that people depend on, it should also > >> > support the latest version of Cassandra, on release. It doesn't look > >> > good > >> > if we release C* without usable drivers or tooling to operate it. It is > >> > massively deflating to announce we just released 5.0 (exciting!) but you > >> > can't use it for 6 months because you're using analytics lib and the > >> > people > >> > that contribute to it has better things to do with their time. I think > >> > part of the obligation of maintaining these projects needs to be keeping > >> > up > >> > with latest releases. If that can't be done, we should continue to treat > >> > it as a use-as-your-own-risk thing without official releases. > >> > > >> > Increasing the size of the ecosystem is challenging... > >> > > >> > Jon > >> > > >> > > >> > > >> > > >> > On Wed, Oct 2, 2024 at 1:21 PM Francisco Guerrero <fran...@apache.org> > >> > wrote: > >> > > >> > > Hi folks, > >> > > > >> > > Thanks for all the input. I'm trying to gather all the comments, and > >> > > from > >> > > what I > >> > > can gather it seems that most of the opinions are converging towards > >> > > scoping > >> > > a Sidecar release. These are the items that were called out that we > >> > > will > >> > > need > >> > > for a release: > >> > > - Documentation > >> > > - Authorization / Authentication > >> > > - vnode support > >> > > > >> > > There are some smaller bug fixes that need to be included that we can > >> > > label > >> > > as part of the 1.0 release.[1][2][3] > >> > > > >> > > If there are any other tasks we need to completed, I encourage the > >> > > community > >> > > to create JIRAs that can be in the release milestone for the Sidecar. > >> > > > >> > > I agree with Stefan that OpenAPI is desirable. OpenAPI is something > >> > > I've > >> > > been > >> > > looking into as well. I'm also glad to see Abhijeet can help with the > >> > > documentation effort, I can also help on that front. > >> > > > >> > > Hopefully, I've captured your comments as truly as possible. > >> > > > >> > > Thanks again for all the feedback. > >> > > > >> > > Best, > >> > > - Francisco > >> > > > >> > > [1] https://issues.apache.org/jira/browse/CASSANDRASC-120 > >> > > [2] https://issues.apache.org/jira/browse/CASSANDRASC-121 > >> > > [3] https://issues.apache.org/jira/browse/CASSANDRASC-122 > >> > > > >> > > On 2024/10/02 15:29:20 Jon Haddad wrote: > >> > > > When I developed some of the original sidecar code, it was based on > >> > > > REST > >> > > > Easy, which would have allowed us to generate the spec > >> > > > automatically. I > >> > > > did this in a similar project. > >> > > > > >> > > > That was removed here: > >> > > > https://issues.apache.org/jira/browse/CASSANDRASC-57 > >> > > > > >> > > > But unfortunately it looks like you can't generate the spec from the > >> > > code. > >> > > > > >> > > > Disappointing, that functionality was really useful. I wish someone > >> > > > had > >> > > > asked me before gutting it. > >> > > > > >> > > > Jon > >> > > > > >> > > > > >> > > > > >> > > > On Wed, Oct 2, 2024 at 11:16 AM Štefan Miklošovič < > >> > > smikloso...@apache.org> > >> > > > wrote: > >> > > > > >> > > > > Something like this: > >> > > > > > >> > > > > https://instaclustr.github.io/instaclustr-icarus-go-client/ > >> > > > > > >> > > > > On Wed, Oct 2, 2024 at 4:54 PM Štefan Miklošovič < > >> > > smikloso...@apache.org> > >> > > > > wrote: > >> > > > > > >> > > > >> While documenting endpoints please use something like OpenAPI > >> > > > >> specification. The sidecar should expose this document itself so > >> > > > >> when > >> > > I go > >> > > > >> to this and that URL, I see all endpoints, I put the payloads / > >> > > parameters > >> > > > >> for them and I can just directly call that, no messing with curl / > >> > > wget or > >> > > > >> programmatically or whatever like that. The barrier to exercise > >> > > > >> the > >> > > basic > >> > > > >> functionality has to virtually not be there. > >> > > > >> > >> > > > >> On Wed, Oct 2, 2024 at 4:13 PM Abhijeet Dubey < > >> > > dubey.abhijee...@gmail.com> > >> > > > >> wrote: > >> > > > >> > >> > > > >>> Hi folks, > >> > > > >>> > >> > > > >>> I have been using Sidecar recently and have found some of its > >> > > > >>> functionalities to be quite useful. Hari and I are also working > >> > > > >>> on > >> > > CEP-40 > >> > > > >>> which aims to introduce live migration features in Sidecar in the > >> > > > >>> near future. > >> > > > >>> > >> > > > >>> However, as others have mentioned, I agree that it currently > >> > > > >>> lacks > >> > > > >>> proper documentation. > >> > > > >>> > >> > > > >>> Since this is an official Apache project, I believe that creating > >> > > > >>> comprehensive documentation would be beneficial. This > >> > > > >>> documentation > >> > > should > >> > > > >>> include an overview, architecture, a list and description of > >> > > > >>> various > >> > > > >>> endpoints, and some examples or tutorials on how to use Sidecar's > >> > > features. > >> > > > >>> > >> > > > >>> This documentation would help people get started with Sidecar and > >> > > lower > >> > > > >>> the entry barrier for many. We can update the documentation > >> > > incrementally > >> > > > >>> as needed, along with future enhancements and new features. > >> > > > >>> However, > >> > > > >>> creating some form of formal documentation would be very helpful. > >> > > > >>> > >> > > > >>> To this end I'm willing and highly interested in writing some > >> > > > >>> form of > >> > > > >>> formal documentation for the Sidecar project. Please let me know > >> > > > >>> your > >> > > > >>> thoughts/opinions on this proposal. > >> > > > >>> > >> > > > >>> > >> > > > >>> > >> > > > >>> On Wed, Oct 2, 2024 at 6:46 PM Štefan Miklošovič < > >> > > smikloso...@apache.org> > >> > > > >>> wrote: > >> > > > >>> > >> > > > >>>> Totally agree with Jon here basically on all fronts. Apache > >> > > Cassandra > >> > > > >>>> Sidecar was always a hard nut to crack for me, that is probably > >> > > > >>>> why > >> > > I have > >> > > > >>>> not been involved with that a lot even that is a great tool to > >> > > > >>>> have > >> > > and be > >> > > > >>>> invested in as I was writing my own sidecar and I found a lot of > >> > > > >>>> similarities and problems Apache's sidecar tries to fix. There > >> > > > >>>> was > >> > > some > >> > > > >>>> invisible barrier I have never managed to jump over. I was > >> > > > >>>> looking > >> > > around > >> > > > >>>> and I am very sorry if I just have not found it yet but there is > >> > > not a list > >> > > > >>>> of endpoints a sidecar has, is there? In readme and dev docs > >> > > > >>>> there > >> > > is just > >> > > > >>>> nothing. Taking it at a face value I just don't know what > >> > > > >>>> Sidecar is > >> > > > >>>> capable of and how to use it. I see in the commit history there > >> > > > >>>> is > >> > > a bunch > >> > > > >>>> of commits mentioning S3 but it is a total blackbox for me as a > >> > > potential > >> > > > >>>> user. > >> > > > >>>> > >> > > > >>>> On Wed, Oct 2, 2024 at 2:52 PM Jon Haddad > >> > > > >>>> <j...@rustyrazorblade.com> > >> > > > >>>> wrote: > >> > > > >>>> > >> > > > >>>>> I don't think we should release sidecar 1.0 without any docs. > >> > > > >>>>> > >> > > > >>>>> I took a look through the closed JIRAs to see what's there. > >> > > > >>>>> Here's > >> > > > >>>>> what I found, please correct me if there's more: > >> > > > >>>>> > >> > > > >>>>> - Lots of stuff related to analytics. > >> > > > >>>>> > >> > > > >>>>> I would be pretty excited for this, but the analytics library > >> > > > >>>>> only > >> > > > >>>>> works with single token clusters. Most folks don't run > >> > > > >>>>> Cassandra > >> > > this > >> > > > >>>>> way. I realize there's some element of everyone needs to > >> > > > >>>>> scratch > >> > > their own > >> > > > >>>>> itch, but I don't think we can really call this a useful > >> > > > >>>>> feature > >> > > if the > >> > > > >>>>> overwhelming majority of folks can't use it. I've worked with > >> > > > >>>>> a > >> > > couple > >> > > > >>>>> hundred teams over the years and can only think of 1 org > >> > > > >>>>> outside > >> > > of Apple > >> > > > >>>>> and Netflix that used 1 token, and It was a cluster that > >> > > > >>>>> predated > >> > > v-nodes. > >> > > > >>>>> > >> > > > >>>>> The analytics repo says it's compatible with Cassandra 4, but > >> > > > >>>>> not > >> > > 5. > >> > > > >>>>> > >> > > > >>>>> - Backup & Restore from S3 > >> > > > >>>>> > >> > > > >>>>> Is this compatible with other cloud providers or object > >> > > > >>>>> stores? It > >> > > > >>>>> specifically lists S3 in JIRA. I haven't looked at the source > >> > > yet. Am I > >> > > > >>>>> correct in reading it supports backing up snapshots, no > >> > > > >>>>> continuous > >> > > > >>>>> backups? Seems like we should have at least feature parity > >> > > > >>>>> with > >> > > Medusa if > >> > > > >>>>> we're going to release something here. > >> > > > >>>>> > >> > > > >>>>> All the other closed JIRAs look related to these two items. > >> > > > >>>>> So the > >> > > > >>>>> question is, are we releasing 1.0 as an limited S3 backup and > >> > > restore > >> > > > >>>>> tool? One that prevents you from upgrading to Cassandra 5 if > >> > > > >>>>> you > >> > > happen to > >> > > > >>>>> use single token clusters? > >> > > > >>>>> > >> > > > >>>>> Who is the target audience? > >> > > > >>>>> > >> > > > >>>>> Jon > >> > > > >>>>> > >> > > > >>>>> > >> > > > >>>>> > >> > > > >>>>> On Wed, Oct 2, 2024 at 2:41 AM Dinesh Joshi <djo...@apache.org> > >> > > wrote: > >> > > > >>>>> > >> > > > >>>>>> Currently the Sidecar has a lot of functionality that is > >> > > immediately > >> > > > >>>>>> usable by the community. Apart from minor fixes, the AuthN/Z > >> > > story would be > >> > > > >>>>>> wrapped up soon. Post this, I would propose moving forward > >> > > > >>>>>> with > >> > > cutting a > >> > > > >>>>>> release with the existing feature set so we can get this in > >> > > > >>>>>> the > >> > > hands of > >> > > > >>>>>> our community. > >> > > > >>>>>> > >> > > > >>>>>> On Tue, Oct 1, 2024 at 8:27 PM guo Maxwell > >> > > > >>>>>> <cclive1...@gmail.com> > >> > > > >>>>>> wrote: > >> > > > >>>>>> > >> > > > >>>>>>> Have the same question ： what ‘s the plan ？ > >> > > > >>>>>>> > >> > > > >>>>>>> Jeff Jirsa <jji...@gmail.com>于2024年10月2日 周三上午10:43写道： > >> > > > >>>>>>> > >> > > > >>>>>>>> > >> > > > >>>>>>>> > >> > > > >>>>>>>> On Oct 1, 2024, at 7:26 PM, Josh McKenzie > >> > > > >>>>>>>> <jmcken...@apache.org > >> > > > > >> > > > >>>>>>>> wrote: > >> > > > >>>>>>>> > >> > > > >>>>>>>> However it is used by a number of other features as a > >> > > > >>>>>>>> dependency > >> > > > >>>>>>>> such as analytics, backup/restore, repair, metrics, and CDC > >> > > > >>>>>>>> > >> > > > >>>>>>>> It seems like a natural pressure relief valve for moving > >> > > operations > >> > > > >>>>>>>> out of a core C* node that are well served out of process. > >> > > > >>>>>>>> > >> > > > >>>>>>>> > >> > > > >>>>>>>> Yea, but the point of the foundation is to RELEASE software > >> > > > >>>>>>>> for > >> > > the > >> > > > >>>>>>>> public good, and the link asserting consensus was dec2018, > >> > > > >>>>>>>> so > >> > > its’ 5.5 > >> > > > >>>>>>>> years and no releases. > >> > > > >>>>>>>> > >> > > > >>>>>>>> What’s the plan here? > >> > > > >>>>>>>> > >> > > > >>>>>>>> > >> > > > >>>>>>>> > >> > > > >>>>>>>> > >> > > > >>>>>>>> > >> > > > >>> > >> > > > >>> -- > >> > > > >>> *Abhijeet* > >> > > > >>> > >> > > > >> > >> > > > > >> > > > >> > >
