+1 to support HBase 2.x But not sure about dropping support for 1.4.x I cc'ed to user@ and user-zh@ to hear more feedback from users. Best, Jark On Thu, 18 Jun 2020 at 21:25, Gyula Fóra <[hidden email]> wrote: Hi All! |
+1 to support HBase 2.x And I think the 1.4.x version can be retained for the time being, so that users who are currently using the 1.4.x version can have more time to evaluate whether their projects need to be upgraded and the cost of upgrading. At 2020-06-19 12:35:36, "Jark Wu" <[hidden email]> wrote:
|
Hi! Having both 1.4.x and 2.x supported means we need different modules or some shim logic as they are not compatible with each other. I would love to avoid this if possible because it is a lot of extra effort from a maintainability perspective. It would be great to see how many users still use HBase 1.4.3 and how many would use a 2.2.x connector instead. Gyula On Fri, Jun 19, 2020 at 7:44 AM chaojianok <[hidden email]> wrote:
|
In reply to this post by Jark Wu-3
+ 1 to support HBase 2.x
and the hbase 2.x client dependencies are simple and clear. Tbe hbase project shades them all
Sent from the Apache Flink User Mailing List archive. mailing list archive at Nabble.com. |
cc'ed YuLi and ZhengHu who are HBase PMC members and may know how many users are using HBase 1.4.x and 2.2.x Best, Jark On Fri, 19 Jun 2020 at 14:20, jackylau <[hidden email]> wrote: + 1 to support HBase 2.x and the hbase 2.x client dependencies are simple and clear. Tbe hbase project shades them all |
In reply to this post by jackylau
+1 to support HBase 2.2.x, and +1 to retain HBase 1.4.3 until we deprecates finished(maybe one version is enough).
Currently we only support HBase 1.4.3 which is pretty old, and I’m making a flink-sql-connector-hbase[1] shaded jar for pure SQL user, the dependencies is a little more complex.
Leonard Xu |
+1 on upgrading the HBase version of the connector, and 1.4.3 is indeed an old version. OTOH, AFAIK there're still quite some 1.x HBase clusters in production. We could also see that the HBase community is still maintaining 1.x release lines (with "stable-1 release" point to 1.4.13) [1] I totally understand the additional efforts of maintaining two modules, but since we're also reserving multiple versions for kafka connector, meantime considering the current HBase in-production status, I'd still suggest to get both 1.4.13 and 2.2.5 supported. Best Regards,
Yu On Fri, 19 Jun 2020 at 14:58, Leonard Xu <[hidden email]> wrote:
|
One supplement: I noticed that there are discussions in HBase ML this March about removing stable-1 pointer and got consensus [1], and will follow up in HBase community about why we didn't take real action. However, this doesn't change my previous statement / stand due to the number of 1.x usages in production. Best Regards,
Yu On Fri, 19 Jun 2020 at 15:54, Yu Li <[hidden email]> wrote:
|
+1 to support both HBase 2.x and Hbase 1.4.x, just as what we are doing for Kafka. On Fri, Jun 19, 2020 at 4:02 PM Yu Li <[hidden email]> wrote:
Best Regards
Michael Li |
Hi According to my observation in the hbase community, there are still lots of hbase users running their production cluster with version 1.x (1.4x or 1.5.x). so I'd like to suggest that supporting both hbase1.x & hbase2.x connector. Thanks. On Sat, Jun 20, 2020 at 2:41 PM Ming Li <[hidden email]> wrote:
|
If we support both HBase 1 and 2, maybe it's a good time to pull them out to Bahir and list them in flink-packages to avoid adding even more modules to Flink core? On Mon, Jun 22, 2020 at 4:05 AM OpenInx <[hidden email]> wrote:
-- Arvid Heise | Senior Java Developer Follow us @VervericaData -- Join Flink Forward - The Apache Flink Conference Stream Processing | Event Driven | Real Time -- Ververica GmbH | Invalidenstrasse 115, 10115 Berlin, Germany -- Ververica GmbHRegistered at Amtsgericht Charlottenburg: HRB 158244 B Managing Directors: Timothy Alexander Steinert, Yip Park Tung Jason, Ji (Toni) Cheng |
If we were to go the bahir route, I don't see the point in migrating the 1.4.x version there since that's already available in Flink. To me that is almost the same as dropping explicit support for 1.4 and telling users to use older connector versions if they wish to keep using it. If we want to keep 1.4 around for legacy users and slowly deprecate that, we can do that inside Flink and only push the 2.4.x version to bahir. What do you think? Gyula On Mon, Jun 22, 2020 at 3:16 PM Arvid Heise <[hidden email]> wrote:
|
I would like to revive this discussion. There's a new JIRA[1] + PR[2] for adding HBase 2 support. it seems that there is demand for a HBase 2 connector, and consensus to do it. The remaining question in this thread seems to be the "how". I would propose to go the other way around as Gyula suggested: We move the legacy connector (1.4x) to bahir and add the new (2.x.x) to Flink. Why? In the Flink repo, we have a pretty solid testing infra, where we also run Hbase end to end tests. This will help us to stabilize the new connector and ensure a good quality. It also, the perception of what goes into Flink, and what into Bahir is a bit clearer if we put the stable, up to date stuff into Flink, and legacy, experimental or unstable connectors into Bahir. Who can take care of this effort? (Decide which Hbase 2 PR to take, review and contribution to Bahir) On Mon, Jun 22, 2020 at 3:32 PM Gyula Fóra <[hidden email]> wrote:
|
It seems that this thead is not on dev@ anymore. Adding it back ... On Fri, Aug 7, 2020 at 9:23 AM Robert Metzger <[hidden email]> wrote:
|
Hi Robert, I completely agree with you on the Bahir based approach. I am happy to help with the contribution on the bahir side, with thorough review and testing. Cheers, Gyula On Fri, 7 Aug 2020 at 09:30, Robert Metzger <[hidden email]> wrote:
|
Hi Robert and Gyula, Thanks for reviving this thread. We have the implementation (currently for 2.2.3) and it is straightforward to contribute it back. Miklos (ccd) has recently written a readme for said version, he would be interested in contributing the upgraded connector back. The latest HBase version is 2.3.0, if we are touching the codebase anyway I would propose to have that. If everyone is comfortable with it I would assign [1] to Miklos with double checking the all functionality that Felipe has proposed is included. On Fri, Aug 7, 2020 at 10:13 AM Gyula Fóra <[hidden email]> wrote:
|
Hi, Thank you for picking this up so quickly. I have no objections regarding all the proposed items. @Gyula: Once the bahir contribution is properly reviewed, ping me if you need somebody to merge it. On Fri, Aug 7, 2020 at 10:43 AM Márton Balassi <[hidden email]> wrote:
|
I'm +1 to add HBase 2.x However, I have some concerns about moving HBase 1.x to Bahir: 1) As discussed above, there are still lots of people using HBase 1.x. 2) Bahir doesn't have the infrastructure to run the existing HBase E2E tests. 3) We also paid lots of effort to provide an uber connector jar for HBase (not yet released), it is helpful to improve the out-of-box experience. My thought is that adding HBase 2.x doesn't have to remove HBase 1.x. It doesn't add too much work to maintain a new version. Keeping the old version can also help us to develop the new one. I would suggest to keep HBase 1.x in the repository for at least one more release. Another idea is that maybe it's a good time to have a "apache/flink-connectors" repository, and move both HBase 1.x and 2.x to it. It would also be a good place to accept the contribution of pulsar connector and other connectors. Best, Jark On Fri, 7 Aug 2020 at 17:54, Robert Metzger <[hidden email]> wrote: Hi, |
Hi all! Im new here; I have been using the flink connector for hbase 1.2, but recently opt to upgrading to hbase 2.1(basically because was bundled in CDH6) it would be nice to add support for hbase 2.x! I found that supporting hbase 1.4.3 and 2.1 needs minimal changes and keeping that in mind last week I sent a PR with a solution supporting 1.4.3/2.1.0 hbase (maybe not the best, im sorry if i break some rules sending the PR). i would be happy to help if needed! Felipe. El 07-08-2020, a la(s) 10:53, Jark Wu <[hidden email]> escribió:
|
@Jark: Thanks for bringing up these concerns. All the problems you've mentioned are "solvable": - uber jar: Bahir could provide a hbase1 uber jar (we could theoretically also add a dependency from flink to bahir and provide the uber jar from Flink) - e2e tests: we know that the connector is stable, as long as we are not adding major changes (or we are moving the respective e2e tests to bahir). On the other hand, I agree with you that supporting multiple versions of a connector is pretty common (see Kafka or elasticsearch), so why can't we allow it for Hbase now? I'm really torn on this and would like to hear more opinions on this. On Fri, Aug 7, 2020 at 11:24 PM Felipe Lolas <[hidden email]> wrote:
|
Free forum by Nabble | Edit this page |