Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    Jordan Zimmerman
    @Randgalt
    @maheswar4b4 Apache Curator supports this
    maheswar4b4
    @maheswar4b4
    @Randgalt Thanks, I will check curator then
    walker
    @lwx2615
    hi,everyone
    walker
    @lwx2615
    the log4j said" Define some default values that can be overridden by system properties",but how to overridden it
    I can't see any log config in the zoo.cfg
    walker
    @lwx2615
    I have got it at other file
    frid
    @FridMehdi
    @hello Folks some one could explain this issue "This ZooKeeper instance is not currently serving requests"
    Jordan Zimmerman
    @Randgalt
    It means your ensemble is not in Quroum. E.g. you have a 3 instance ensemble and 2 instances are down
    enixon
    @enixon
    @lwx2615 you can override logging properties by passing java args into the process, for example '-Dlog4j.configuration=' can be used to specify a custom logging config file. This cannot be set through the zookeeper configuration file.
    @FridMehdi any instance responding with "not currently serving requests" is not connected to the ensemble leader (possibly because there isn't one but also if there's some network partition).
    walker
    @lwx2615
    @enixon thank you!
    namnhcntt
    @namnhcntt
    Hi everyone, I faced issue with zookeeper in docker: "Refusing session request for client /10.255.0.2:42078 as it has seen zxid 0x1b0000001b our last zxid is 0x1b0000001a client must try another server". Please help me resolve this issue, thanks
    image.png
    enixon
    @enixon
    @namnhcntt , you'd see this error message if a client has disconnected from an active zookeeper server and is attempting to reconnect to another server in the ensemble that is lagging behind. I'd expect that the session timeout at the client (how long it's disconnected before trying the next server) would be more than any server lag you might be seeing in the quorum but these are both avenues to explore. Try seeing if any of your servers are experiencing extensive lag compared to the others or if your session timeout on the client is set too low.
    yctn
    @yctn
    i rebooted one of my zookeeper nodes but it refuses to join. so i restart all nodes and now none are able to restart. all show me this error Cannot open channel to 3 at election address
    joshblease
    @joshblease
    Hello all! I'm trying to repair a bookkeeper cluster with the auto-recovery tool but it doesn't want to run along side the current process - " Address already in use"
    Looks like a conflict for prometheusStatsHttpPort
    joshblease
    @joshblease
    Right, changed that port and we're running. The auto recovery is attempting to connect to a bookie which has been lost (VM issues), is this expected?
    Enrico Olivelli
    @eolivelli
    @joshblease welcome to this channel, I am from BK community. Can you please move this question to BK slack channel? You can find it on bookkeeper.apache.org
    rajatgangwar
    @rajatgangwar

    [curator][lock-recipe] pre ZK 3.5 (container nodes), curator lock implementation creates persistent parent node and ephemeral-sequential child node. I believe this is done to mitigate herd-effect. But we have a use-case where 99.99% calls are unique, and we need lock just in case 2 duplicate calls come at the same time. Given that we are still stuck with ZK 3.4 and I want to avoid reaper (since it is deprecated), can I implement a lock recipe which just tries to create ephemeral node without any persistent nodes. Since ZK will fail if node already exists, this will ensure only one call (among the duplicate concurrent calls) will go through ?

    Are there any corner cases where it will not work ?
    Since I didn't find any library(recipe) with such implementation I was not sure about it.
    Can someone please tell me if there is any problem with this approach for the use-case where duplicates are very rare, but whenever it happens we need to make sure only one call goes through?

    Ajith Kumar Rajendran
    @AjithKumar-Raj
    Is one zookeeper Ensemble can handle multiple kafka cluster?
    Sahaay Platform In Progress, more coming...
    @Ekshunya-India
    Hello Team, I am completely new to zookeeper. I am currently going through O'reilley zookeeper book. I have a couple of general question about zookeeper. 1. Is it a good or a Bad idea to expose our zookeeper instance to public with authentication ofcourse? 2.If it is good to expose zookeeper to public, then is it a good idea to have mobile devices connect to the Zookeeper as clients? Any inputs is much appreciated. Thanks.
    Sahaay Platform In Progress, more coming...
    @Ekshunya-India

    Hello Team, I am completely new to zookeeper. I am currently going through O'reilley zookeeper book. I have a couple of general question about zookeeper. 1. Is it a good or a Bad idea to expose our zookeeper instance to public with authentication ofcourse? 2.If it is good to expose zookeeper to public, then is it a good idea to have mobile devices connect to the Zookeeper as clients? Any inputs is much appreciated. Thanks.

    Actually the more i read, i understood that it is always good to have zookeeper behind a load balancer. And i hope the load balancer can handle authentication. Thanks.

    Kindly let me know if my understanding is wrong.
    RAJKUMAR NATARAJAN
    @rajcspsg
    Hi
    nicolas lecart
    @nlecart_twitter
    Hi all, I am looking for a C++ client library for Zookeeper. I just can find out one library with just few contributors and poor activities. Anybody with experience with C or C++ Zookeeper client? thanks !
    enixon
    @enixon
    @nlecart_twitter The client I'm most familiar with is the C client maintained by the Apache ZooKeeper community directly (https://zookeeper.apache.org/doc/current/zookeeperProgrammers.html#C+Binding). I don't know of a public C++ client proper.
    jalindar96
    @jalindar96
    Hi All,
    I am getting below error while starting with zkCli.
    I am having ZK 3.5.9 & curator framework 3.3.0
    org.apache.zookeeper.KeeperException$NoNodeException: KeeperErrorCode = NoNode for /servi, I am accessing path /services/cdp-jobs, (Want to get payload of spark streaming job)
    Matt
    @pabrahamusa
    Hi guys, trying get head around a memory related issue with zookeeper. I have 5 zookeeper instances running and the memory usage increases slowly and finally it crashes. I updated Xmx to 2048 and max memory for the Pod to 8GB. However it is still growing. Is this a known issue?
    Andor Molnár
    @anmolnar
    @pabrahamusa Be aware that the entire ZK database is stored in memory, so memory usage depends on your usage pattern. If you keep adding data to Zookeeper, the memory usage will increase. Btw it’s better to join the ASF slack org and ask ZK questions there or at user@ list, because this channel is not actively monitored.
    Didip Kerabat
    @didip
    Hello folks, I have a small question if you don't mind. Have you ever seen a case where you are unable to delete some ZK records? rmr and delete doesn't work. What can I do to force delete records?
    mihdih
    @mihdih
    Hello! Good day everyone! I am currently playing around with zookeeper, version 3.5.8 and ran it under a docker container. I started with 3 instances to form the ensemble. All are great and everything work as expected. However when I tried to increase the ensemble to 5 instances. Once the new instances became the leader it suddenly stop accepting queries. My configuration looks correct, where in the 2 new instances are added. Is there any other configuration (maybe manual) that i need to apply?
    mihdih
    @mihdih
    ah please ignore my last message, looks like there was a bug in my setup
    anyhow, another question. is it possible to set a preferred leader? let say we have 3 instances in ensemble, zk1, zk2, zk3. zk3 being the current leader. Is there a way to assign zk1 as leader without stopping zk3?
    enixon
    @enixon
    @mihdih there's no current way to dynamically assign the leader. I remember working on a utility that might do exactly that a while back and thinking it would require modifying the leader election protocol to pull it off.
    Anders Chen
    @chen-anders

    hi I've been having a lot of trouble trying to get a working bcfks keystore for zookeeper TLS ; getting a "DER length more than 4 bytes: 109" error; been debugging for a few hours and nothing quite sticks out

    script/zk settings:
    https://gist.github.com/chen-anders/f802f836454ddbd5c5ee52178c9eb5d8

    wondering if I missed something obvious here

    Anders Chen
    @chen-anders
    turns out setting keystore.type=BCFKS to keystore.type=FIPS fixed it
    Rammohan Ganapavarapu
    @rammohanganap
    Hi, does zookeeper support reads out of observers and or followers while leader is down ?
    enixon
    @enixon
    @rammohanganap , since 3.4 you can set the property readonlymode.enabled to allow voting peers and observers to serve reads while they cannot see the current leader (or no leader has been elected). The property defaults to false if not set. More reading at: https://zookeeper.apache.org/doc/r3.8.0/zookeeperAdmin.html#Experimental+Options%2FFeatures
    Rammohan Ganapavarapu
    @rammohanganap
    @enixon thank for pointing me to the doc readonlymode.enabled means when leader is available they accept write and serve reads but when leader is not available then they can still serve reads right ? Any changes on client side (i assume no)?
    enixon
    @enixon
    That understanding is correct. There's no need to change anything on the client side to take advantage of the feature but you may want to check your reconnect timings as I believe at one point the feature required a client reconnect when switching from read-write to read-only and again when switching back from read-only to read-write. It may have been improved since then.
    And the usual caveat that you'll have to independently monitor for partition in your ensemble. Clients could get stuck on a read-only server if it is partitioned from the rest of the servers unless you actively work to move their connections.
    Rammohan Ganapavarapu
    @rammohanganap
    I see, thank you. will test it out.
    Saurav Deb
    @SauravDeb
    Hi folks. I'm trying to get a ZK 3.6.3 cluster running and using Exhibitor as the supervisor. I have 3 nodes; each having its own Exhibitor process which in turn kicks off the ZK; it looks like the exhibitors are able to talk to each other but as soon as one server IP added to the cluster list, the previous entry disappears; meaning only 1 ZK node is able to stay up at a time, with the others crashing down with the typical MyId file missing error. Looks like some kind of a race condition is going on and only one node is able to register itself on the exhibitor cluster list. What exactly is happening here?
    Sruthi P. C
    @pcsruthi00_gitlab
    Hi all, are zk transaction logs stored in mem as well as on disk? Could an extremely large transaction log prompt gc?
    enixon
    @enixon
    ZooKeeper does not hold all the data for any specific transaction log file in memory during standard operation (i.e. while it's writing to files and not reading from files). To my knowledge there is nothing held in memory that would grow in size along with a transaction log or need to be cleaned up specifically when rolling transaction logs.
    Sruthi P. C
    @pcsruthi00_gitlab
    Got it, thank you!
    Malthe Borch
    @malthe
    This issue is marked (correctly in my opinion) as CRITICAL and hasn't seen any resolution (initially reported May, 2020): https://issues.apache.org/jira/browse/ZOOKEEPER-3822. Since 3.5.x series is EOL since 1st of June, 2022, there is actually no current release which doesn't have this problem (it would seem).
    Akhil Karanth
    @akhil-karanth
    :wave: We recently upgraded our zookeeper version to 3.6.3. I am in process of upgrading the java client/curator recipes.
    I noticed that TreeCache class is deprecated in favor of CuratorCache. We use a TreeCacheSelector with the TreeCache to watch only a part/branch of the zk tree. Is there similar i can use with CuratorCache ?