[09:32:54] <elukey>	 https://engineering.linkedin.com/blog/2021/the-exabyte-club--linkedin-s-journey-of-scaling-the-hadoop-distr
[09:33:04] <elukey>	 "LinkedIn now stores 1 exabyte of total data across all Hadoop clusters."
[09:33:13] <elukey>	 :O :O :O
[09:33:59] <elukey>	 this blog post is great, it contains a ton of useful infos
[09:34:02] <elukey>	 "since the NameNode startup can take up to an hour"
[09:34:03] <elukey>	 hahahahah
[09:34:46] <elukey>	 they talk about doing rolling upgrades, I think that we should think about it now that we are on 2.10
[09:35:12] <elukey>	 the most impressive thing is
[09:35:14] <elukey>	 "As of this writing our largest NameNode is set to use a 380 GB heap to maintain 1.1B of namespace objects"
[09:44:02] <elukey>	 the fair vs non-fair locking is also nice, but I am wondering if they patched the code or used a jvm setting (I think the former)
[09:44:12] * elukey waits for Joseph to be nerd-sniped by this article
[09:46:03] <elukey>	 wow https://issues.apache.org/jira/browse/HDFS-12943 says that they backported it to 2.10
[10:10:52] <wikibugs>	 10Analytics: hdfs dfsadmin saveNamespace fails on an-master1001 - https://phabricator.wikimedia.org/T283733 (10elukey) I took a look to the Threads dump in the HDFS Namenode logs right after the issue with `saveNamespace` after a conversation between me and Joseph about the number of service handler threads (the...
[10:11:00] <elukey>	 joal: --^
[10:11:02] <elukey>	 super interesting
[18:45:17] <joal>	 elukey: hehehe - my aim was to try to nerd-snipe you after weekend :D Let's talk about all that next Tuesday :)