Sure thing. Context: I'm interested in service discovery solutions (e.g. SmartStack). (Caution: loose brainstorming ahead)
Yes, I suppose I'm thinking about how they sync to one another across geography and network partitions. If I understand correctly, one of many great things about etcd over a traditional LDAP store is its chosen method of sync (RE: Two Generals' problem, they sync via Raft instead of Paxos(?)). Without botching the facts, I'll just say that after I read last year about Raft vs. Paxos, I was impressed but was forced to shelve my curiosity for another day.
To some extent, I sense a yet-to-be created but potentially very useful/marketable (open source or otherwise) product in this space somewhere (LDAP vs. etcd/Zookeeper/etc, Service Discovery and configuration management). I should definitely go back and read more about Raft and Paxos.
In the mean time, thinking more about the problem, I am also not certain how Master-Master is handled in most LDAP sync'ing schemes, though I suspect it would be Paxos.
I suppose my question is somewhat of a proxy to this one:
I am pondering seeding an LDAP cluster (non-traditional schema) from etcd (or similar). A scalable and reliable source of truth is what I'm looking for, I suppose. Would that gain me anything in the way of strict ACLs on reads? etcd does not appear "enterprise ready" in this regard. I'll have to do more research to know. My mind plays crazy tricks sometimes when reading about new tech on HN and lamenting some old tech on the job(s), so there's little accounting for that.
Another off the wall thought I had was to evaluate for value creation of an etcd-like service starting from SpiderOak's Crypton framework. In a phrase: A zero-knowledge source of truth for service discovery.
2 comments
[ 3.0 ms ] story [ 17.8 ms ] threadYes, I suppose I'm thinking about how they sync to one another across geography and network partitions. If I understand correctly, one of many great things about etcd over a traditional LDAP store is its chosen method of sync (RE: Two Generals' problem, they sync via Raft instead of Paxos(?)). Without botching the facts, I'll just say that after I read last year about Raft vs. Paxos, I was impressed but was forced to shelve my curiosity for another day.
To some extent, I sense a yet-to-be created but potentially very useful/marketable (open source or otherwise) product in this space somewhere (LDAP vs. etcd/Zookeeper/etc, Service Discovery and configuration management). I should definitely go back and read more about Raft and Paxos.
In the mean time, thinking more about the problem, I am also not certain how Master-Master is handled in most LDAP sync'ing schemes, though I suspect it would be Paxos.
I suppose my question is somewhat of a proxy to this one:
https://news.ycombinator.com/item?id=9014319
I am pondering seeding an LDAP cluster (non-traditional schema) from etcd (or similar). A scalable and reliable source of truth is what I'm looking for, I suppose. Would that gain me anything in the way of strict ACLs on reads? etcd does not appear "enterprise ready" in this regard. I'll have to do more research to know. My mind plays crazy tricks sometimes when reading about new tech on HN and lamenting some old tech on the job(s), so there's little accounting for that.
Another off the wall thought I had was to evaluate for value creation of an etcd-like service starting from SpiderOak's Crypton framework. In a phrase: A zero-knowledge source of truth for service discovery.
https://github.com/SpiderOak/crypton
/end layman's rant now. :-) I think too much and research/build too little. Errr!