Hello, Daan, Rafael. Ok, let's discover some points.
Rafael: All the stuff is managed in Elasticsearch, clustered ES is supported as well. As for now, value datatype is 'keyword' and doesn't fit well for blobs. But, actually, there are no practical limitations except we don't want press the CS server with streaming for blobs as it's design is not well designed for long lasting jobs like that. To adress the case we also plan to develop artifact storage which is intended to keep blobs with versioning and managed security. Probably will use Ceph under the hood. Daan: Let's go with the cases. Actually, in the past I published the announcement for log management plugin which intention is to provide access to VM logs from the API and other interfaces. Now we develop CLI utility for it. That plugin helps to observe what logs are generate during VM deployment. This plugin (KV) helps to deploy complex topologies, like Docker Swarm, Kubernetes, MySQL Galera replicaset, DC/OS, monitor the state of the deployments and do trivial things: Observe current VM load, RAM availability, disk space. All these parameters can be observed in historical perspective or last values. A quorum example. Say, you deploy two VMs programmatically and they have to decide who is Docker master and who is worker, so you create shared storage and both VMs place they UUID in 'order' key. They check the history for 'order' and find who is the first. Next, the first generates cluster initialization and places secret key to 'secret'. After, worker reads it and joins the cluster. Want to add a new VM to the cluster, just pass that KV to it as well. Of cource, proper templates with bootstrap code are required. Actually, there are tons of very sophisticated deployments can be implemented in a very generic way with theese two plugins. Another piece of software to be developed is one-time accessor plugin which is important for secure sharing of various credentials when multi tenant VMs are used. Have a good day. пт, 21 сент. 2018 г., 2:21 Rafael Weingärtner <rafaelweingart...@gmail.com>: > That is very interesting Ivan! > > I have a few questions though: > > - Do you have a limit on ”value” size? Can it be a binary of a file? > - Where do you guys store these “keys” and ”values”? > > > On Wed, Sep 19, 2018 at 4:02 AM, Ivan Kudryavtsev < > kudryavtsev...@bw-sw.com> > wrote: > > > Hello, community. > > > > We glad to announce a new Open Source plug-in for Apache CloudStack which > > helps to maintain generic key-value storages. There are three types are > > supported right now: > > > > - temporary KV with TTL (many per account); > > - persistent KV bound to VM (one per VM); > > - persistent KV bound to account (many per account). > > > > KV storages support a commit-log based approach where old values are not > > erased with new ones and it provides the access to the last key/values > and > > to the history which helps to maintain things like quorums and other > > operations which require serialized access. > > > > The plugin uses Elasticsearch. I > > > > If you are interested to know more, contact our website for the full > > announcement and download details: > > https://bitworks.software/en/2018-09-19-acs-vm-kv-storage-plugin.html > > > > All questions, improvements and considerations are welcome! > > > > -- > > With best regards, Ivan Kudryavtsev > > Bitworks LLC > > Cell: +7-923-414-1515 > > WWW: http://bitworks.software/ <http://bw-sw.com/> > > > > > > -- > Rafael Weingärtner >
