Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    asda777
    @asda777

    Hi, I'm trying to use Olric in embedded mode over Kubernetes (google Kubernetes engine to be specific) and the pods fail to discover each other and coordinate. I tried using the default local/lan/wan envs but when printing stats it seems all partitions are pointing only to the pod's own IP.

    I set up olric-debug and connected to the pods running the embedded Olric. It seems that running olric-cli send you to one of the pods randomly and send your inputs only to it. but the pods still do not share this data.
    Also, setting up olricd cluster (kubectl apply -f https://raw.githubusercontent.com/buraksezer/olric-kubernetes/master/olricd.yaml), connecting to it and running olric-stats will show that all partitions belong to the same IP (alternating ips on each run of the command)

    I also tried to set Olric conf serviceDiscovery property but got an error that "I didn't reach cluster quorum" which I assume means I did it wrong and it didn't set up even the pod own instance correctly.
    Is there an example for an embedded Kubernetes setup/do you have any pointers to how to set this up correctly?
    Thanks in advance.

    Burak Sezer
    @buraksezer
    @asda777 hi. I'm so sorry for the late answer. I'm going to investigate the problem. Thank you for reporting. You may want to create an issue on GitHub for fast response.
    Steven Wade
    @suederade

    Hey guys, I'm trying to get Olric to work in AWS environment and running into some pretty substantial issues. Because of our deployment process we deploy-one then deploy-all. The first node loads a large amount of data, then a second node will come up and successfully attach (using the consul plugin). I can query the data from the new node that is set as the backup and it successfully gets the data, although it seems to emit hundreds of thousands of errors [ERROR] Failed to call get on a replica owner: <ip>: key not found.

    Another problem that happens is when I scale up the instances (e.g. 2 instances to 4), I basically get catastrophic errors that kill the service. It always seems to happen the same way.

    First error is operation timeout from creating a new DMap.

    Second, gob won't register the type gob: name not registered for interface: "*models.Product".

    Third we get a massive error (that gets cut off, sorry) when calling get dial tcp <ip>:8080: connect: cannot assign requested address: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal server error: internal se The nodes seem to think that they connect just find according to the beginning of the logs.

    I have it set up with 120 partitions, replica count of 2.

    Steven Wade
    @suederade
    I may need to look at something else because the error from creating a DMap should actually be exiting the entire program
    Steven Wade
    @suederade
    (Update, it was)
    Steven Wade
    @suederade
    I'm also getting write quorum not reached errors with a quorum count of 1
    Saeid
    @saeidakbari
    Hello everyone and thanks for developers and maintainers who are walking on this project.
    I have a simple questions that maybe sounds ridiculous at first but it's getting important when you have a large amount of data. My question is, does Olric distribute data at the level of DMaps or each individual keys?? I have a disability map that could potentially have a lot of data. if it seems to data at the level of distributed map each time I'm it should sync a large size of data among the network and it would kill the performance.
    Alaa Qutaish
    @caspereeko_twitter
    Hi, I am trying to find architecture diagram or document for Olric, I am very interested in understanding the underlying theory behind its replication protocol. Thank you for your support.
    jasonbronson
    @jasonbronson
    Does the client configuration require a localhost bind for kubernetes or the actual dns entry? We keep getting errors about binding
    Burak Sezer
    @buraksezer
    @jasonbronson hi. I need more information about your case to make a useful comment. But Olric needs actual dns entry to function in prod. If you set 0.0.0.0 as BindAddr, the server tries to discover the IP address. The client needs an IP address or dns entry. Could you please open an issue for this?
    @caspereeko_twitter, I'm so sorry for this extremely late response. Olric follows CAP and PACELC principles. I had shared some documents about the theory in readme document. I found this video very useful: https://www.youtube.com/watch?v=7zU9b48GUGQ