search

blt / locker

An Erlang/OTP per-cluster lock application.
MIT License
8 stars 2 forks source link

Should locker be tolerant of node failures? #11

Open blt opened 11 years ago

blt commented 11 years ago

As of 54ba0a82e332a2c8db04fad413dc9199cd2b5765 Locker achieves per-cluster locks by registering a global application. Locks are held on a single node . In the event of failure of this node, Locker can be restarted on another node in the cluster, but all locks state will be lost.

I am undecided if this is a problem. It's almost certainly not, truly, for my use-case.

msassak commented 11 years ago

Using mnesia to keep state synchronized across all the nodes seems like a good idea. Also depending on the number of locks, using write rather than dirty_write might work okay.

blt commented 11 years ago

@msassak Been thinking about how it'd look if riak_core was used; would have to identify the tradeoffs of the two approaches to make a better decision.

msassak commented 11 years ago

IMHO a dependency like riak/core seems overkill at this stage, but I have no experience with it.

On Mon, Aug 19, 2013 at 11:48 AM, Brian L. Troutwine notifications@github.com wrote:

@msassak Been thinking about how it'd look if riak_core was used; would have to identify the tradeoffs of the two approaches to make a better decision.

Reply to this email directly or view it on GitHub: https://github.com/blt/locker/issues/11#issuecomment-22895043

blt commented 11 years ago

That is a point in mnesia's favor: no external dependency.

On Mon, Aug 19, 2013 at 12:10 PM, Mike Sassak notifications@github.com wrote:

IMHO a dependency like riak/core seems overkill at this stage, but I have no experience with it. On Mon, Aug 19, 2013 at 11:48 AM, Brian L. Troutwine notifications@github.com wrote:

@msassak Been thinking about how it'd look if riak_core was used; would have to identify the tradeoffs of the two approaches to make a better decision.

Reply to this email directly or view it on GitHub:

https://github.com/blt/locker/issues/11#issuecomment-22895043

Reply to this email directly or view it on GitHub: https://github.com/blt/locker/issues/11#issuecomment-22896675