Live data from Hacker News

Elasticsearch is dead, long live Open Distro for Elasticsearch

logicalclocks.com

51–60 of 60 posts

Re: Elasticsearch is dead, long live Open Distro for Elasticsearch

#51
post #38
post #12

Earlier quoted context omitted.

My PoV: a bunch of startups chose to do open source because it’s good PR and marketing. Then it proved to be bad for their business model so they moved to something much closer to shared source. Personally, I don’t think there’s anything immoral or shameful about selling proprietary software or shared source or whatever. However, SSPL and TSL and friends are just not open source licenses. They’re more restrictive tha…

"Bad for their business model" because Amazon has tremendous power. It wasn't just a series of bad choices by the startups that got them here, it was concentration of an emerging market over 10-15 years. Amazon dominates conversations because they extract the most value from open source projects. They make more money selling PostgreSQL, MySQL, Elasticsearch, _Docker_, Linux, etc than the creators of those projects. I…

This is exactly why I convinced my company to buy Airflow from Astronomer and Presto from Starburst and Spark from Databricks instead of getting all that from AWS.

I would rather give the money to the people who make the stuff than to people who (all but) steal the stuff and sell it to others.

And it usually comes out cheaper, too!

Re: Elasticsearch is dead, long live Open Distro for Elasticsearch

#52
post #49
post #38

Earlier quoted context omitted.

"Bad for their business model" because Amazon has tremendous power. It wasn't just a series of bad choices by the startups that got them here, it was concentration of an emerging market over 10-15 years. Amazon dominates conversations because they extract the most value from open source projects. They make more money selling PostgreSQL, MySQL, Elasticsearch, _Docker_, Linux, etc than the creators of those projects. I…

> They make more money selling PostgreSQL, MySQL, Elasticsearch, _Docker_, Linux, etc than the creators of those projects. It's not the only perspective, but it dominates for a reason. They make more money selling EC2, with this software installed on top of it. If Amazon did not sell RDS, their customers would buy EC2 instances and install it on there. Which of these realities would you prefer: 1. Mongo, ES, PG, etc…

I feel like [3] is the heart of the matter. The value is open-source to companies is that it commodifies support. Also software at an enterprise scale costs money. Not being held hostage allows you to switch support providers or go internal or fly blind.

What these new quasi-open source (ie source available) companies want to do leverage their authorship of the software to be sole support service provider (of meaningful size)... which is exactly the sole value the enterprises extract from opensource: the lack of vendor leverage!

The market does not tolerate a contradiction.

Re: Elasticsearch is dead, long live Open Distro for Elasticsearch

#53
post #52
post #49

Earlier quoted context omitted.

> They make more money selling PostgreSQL, MySQL, Elasticsearch, _Docker_, Linux, etc than the creators of those projects. It's not the only perspective, but it dominates for a reason. They make more money selling EC2, with this software installed on top of it. If Amazon did not sell RDS, their customers would buy EC2 instances and install it on there. Which of these realities would you prefer: 1. Mongo, ES, PG, etc…

I feel like [3] is the heart of the matter. The value is open-source to companies is that it commodifies support. Also software at an enterprise scale costs money. Not being held hostage allows you to switch support providers or go internal or fly blind. What these new quasi-open source (ie source available) companies want to do leverage their authorship of the software to be sole support service provider (of meaning…

Exactly. Companies like Mongo, ES, etc leveraged open source and venture capital to get where they are today, and now they're finding that open source isn't the business that venture capital wants it to be.

I really don't know why they get so much defense in the community; I view their actions as a straight betrayal. The one thing you can say about AWS is, they're the enemy you know. They don't have a history of turning their back on OSS, but they also don't have a history of abusing OSS to gain traction for their product (though, the MySQL/Aurora relationship could reasonably fall under this. I think its different; not a betrayal because they haven't turned their backs on MySQL, but are just trying to make it better for their customers in ways MySQL alone can't do). With AWS, you know what you get; they care about the tech and the money; no politics (well, almost), no weird philosophies guiding decision making, just solving problems and getting paid to solve those problems. That's infrastructure, baby.

Its a lot hazier with, say, MongoDB. You wouldn't believe the pressure they've put on customers on 3.6; we're talking weekly emails, even a few cold calls, from sales reps. "Hey, I know its the holidays, but have you taken a look at 4.0 yet? Let me send you a link to check it out, lots of cool stuff." Yeah; the real reason they want you to upgrade is because 4.0 has a bunch of new features that competitors like AWS can't replicate due to its new licensing. I'd love to hear from someone running MySQL on RDS, and if they've had a similar experience with AWS sales concerning Aurora; I suspect they haven't. Frankly, AWS never talks to us and we never talk to them, unless there's a problem or we want a discount, and they're always accommodating for both. The perfect relationship.

Re: Elasticsearch is dead, long live Open Distro for Elasticsearch

#54

Not too long ago, I tried to install the ELK stack to enable me to ingest some log files and easily view them on a web page. After many hours of digging through documentation, I gave up. The process was a train wreck of obscurity, complexity and heavy weight processes, just to do a relatively simple task. I've been doing open source for 20+ years now and one thing that we still haven't seemed to get right is: a simpl…

I've been a mostly happy user of Elasticsearch for 5+ years, but agree it is a bit heavy for just logs (which is ironic, since that's the use case that really launched it beyond application search). I recommend Loki for logs, especially if you're already using Grafana: https://grafana.com/oss/loki/

Re: Elasticsearch is dead, long live Open Distro for Elasticsearch

#55
post #10

PSA: This site is basically an ad. And not one about Elasticsearch at that. I don't understand what all the noise is about. The new license does not make ES closed source, as far as I can tell. They re-license under the MongoDB server side license, which reads a lot like a rephrased AGPL to me. None of you had any reasonable expectation of getting a free Elasticsearch to begin with, and decades worth of free maintena…

At the end of the day this is Copy-Left debate all over again

Personally I like it, I see "Open Source" being killed off with Apache, BSD. MIT and other non-copy left licensing, it allows companies like Google, Apple, Amazon, Microsoft and other to leach off of "Open Source" focusing exclusively on dev tools and libraries used to build their products.

This has lead to ALOT of developer growth, and ALOT of Open Source access for Developers but no much growth in END USER software products that are open sourced.

Linux one of the most popular Operating Systems in the world, and is the final product not a development library is as big as it is because of Copy Left, one can make a big case that the BSD's never got to be as big as Linux directly because of their permissive license.

Re: Elasticsearch is dead, long live Open Distro for Elasticsearch

#56
post #54

Not too long ago, I tried to install the ELK stack to enable me to ingest some log files and easily view them on a web page. After many hours of digging through documentation, I gave up. The process was a train wreck of obscurity, complexity and heavy weight processes, just to do a relatively simple task. I've been doing open source for 20+ years now and one thing that we still haven't seemed to get right is: a simpl…

I've been a mostly happy user of Elasticsearch for 5+ years, but agree it is a bit heavy for just logs (which is ironic, since that's the use case that really launched it beyond application search). I recommend Loki for logs, especially if you're already using Grafana: https://grafana.com/oss/loki/

Years ago, ELK was the thing for log files. logstash. https://logz.io/blog/server-log-analysis/

loki is indeed what I ended up with and it was a mostly easier install process (despite the missing easily installed debian packages).

The next problem being that Graphana is actually quite terrible at rendering log files. It seems the log file usecase just isn't that popular.

Re: Elasticsearch is dead, long live Open Distro for Elasticsearch

#57
post #55
post #10

PSA: This site is basically an ad. And not one about Elasticsearch at that. I don't understand what all the noise is about. The new license does not make ES closed source, as far as I can tell. They re-license under the MongoDB server side license, which reads a lot like a rephrased AGPL to me. None of you had any reasonable expectation of getting a free Elasticsearch to begin with, and decades worth of free maintena…

At the end of the day this is Copy-Left debate all over again Personally I like it, I see "Open Source" being killed off with Apache, BSD. MIT and other non-copy left licensing, it allows companies like Google, Apple, Amazon, Microsoft and other to leach off of "Open Source" focusing exclusively on dev tools and libraries used to build their products. This has lead to ALOT of developer growth, and ALOT of Open Source…

Indeed, the proof being how much they get back from Apple, Sony.

In fact, had Linux not been around, most likely all major UNIXes would be still be around, using pieces of BSD as always.

Re: Elasticsearch is dead, long live Open Distro for Elasticsearch

#58
post #5

Rephrased: Amazon played hardball with open source Elasticsearch, limiting elastic's ability to make money and forcing a license change. Now you should use Amazon's fork instead! This is not a good outcome for people who want an independent OSS ecosystem.

> This is not a good outcome for people who want an independent OSS ecosystem. I used to manage a licensed ELK cluster. The licensing was expensive (order of six figures) and was really only necessary for LDAP integration. Features like timelion were cool, but definitely not worth the price tag. I migrated to an AWS-managed solution to for two reasons: 1) to alleviate the burden of managing a cluster, 2) get away fro…

> not having to deal with sales people.

This aspect does not get nearly the attention it deserves.

Unless your product is an exceptionally well focused developer tool, it will take months, and sometime multiple quarters, to see whether it adds any real value. Exceptions exist but are rare. In practice the generous[TM] 30-day trial period is almost always worthless, and often a gateway to sustained net negative.

Add to that the constant barrage of dark patterny attempts to infest your calendar and the upsell pressure tactics. Every third-party integration is an additional liability for maintenance, and far too frequently an increase in attack surface. Time is your most precious asset and the sales people are trying to rob you of it.

To every sales person on the planet: if you want to sell me a service or product, then sell me NOTHING BUT the service or product. Make useful documentation accessible, and make additional features discoverable, with unambiguous pricing clearly in place where necessary. Then get out of my way.

With OSS product I can install and trial it as long as I need, without needing to give off my contact details or be added to your invasive CRM. If I need additional features that are only available under your more restrictive license offering, I will buy them when I have the need. Your sales quota or bonus thresholds do not factor in, and if you try make them my problem, I will look at migration options.

Re: Elasticsearch is dead, long live Open Distro for Elasticsearch

#59
post #22
post #10

PSA: This site is basically an ad. And not one about Elasticsearch at that. I don't understand what all the noise is about. The new license does not make ES closed source, as far as I can tell. They re-license under the MongoDB server side license, which reads a lot like a rephrased AGPL to me. None of you had any reasonable expectation of getting a free Elasticsearch to begin with, and decades worth of free maintena…

> None of you had any reasonable expectation of getting a free Elasticsearch to begin with That is sure what "Apache 2.0: now and forever", as they previously promised, means to me.

Me too long ago.

Nowadays I'd instead interpret that as "security updates for the current version forever",

... after so many smaller companies having gone SSPL or something like that

Re: Elasticsearch is dead, long live Open Distro for Elasticsearch

#60
post #5

Rephrased: Amazon played hardball with open source Elasticsearch, limiting elastic's ability to make money and forcing a license change. Now you should use Amazon's fork instead! This is not a good outcome for people who want an independent OSS ecosystem.

By fork they mean: a binary build straight from the unmodified elastic source code using the elastic build file that creates the OSS build bundled with a handful of plugins and components to add alternatives some of the things elastic bundles. They don't even build it themselves and it gets downloaded in binary form straight from Elastic's servers. If you don't believe me, this is the line in the code where they 'for…

Disclosure: I work for Amazon, but not directly on search related technology.

Some information about Amazon's investments in upstream Apache Lucene can be found here: https://aws.amazon.com/blogs/opensource/amazon-giving-back-a...

Covered in the 2019 blog post are:

  * Concurrent updates and deletes
  * Indexing custom term frequencies
  * Lazy loading of Finite State Transducers
Post reply on HN