I need to upgrade our redis cluster(s) from v3.2.12 to latest (v6). What would be the easiest route for upgrading?
We run Redis as well as Sentinel on each node.
3 node cluster -> 1 master, 2 slaves
Thank you very much
The upgrade is pretty standard. i would suggest to upgrade in this order.
First upgrade the sentinel nodes which are currently are slaves/replica nodes.
After upgrade is completed, it will sync from master node, note that a node can sync from version 3 to version 6 node but not vice-versa.( You can't go back to earlier version)
Use sentinel to Failover to upgraded node.
Upgrade the old master node.
Related
Currently I have a single DC cluster with 3 nodes running 4.1.7 version of Scylla. This setup has been running for a long time and I don't want to make changes to this DC, if possible. Now I have a requirement to add another DC cluster with 3 nodes. Can I set up this new DC with the latest stable version of Scylla? Will the two DCs be able to communicate with each other without any issues? Or am I forced to upgrade the existing DC to the latest version?
Scylla supports rolling upgrades, which means you can indeed upgrade just some of the nodes in the cluster while the rest are still running the older version. The cluster should be able to fully work in this state - including the communication between old and new nodes. Not all upgrade paths are equally supported or have been equally tested, obviously, but most "interesting" upgrade paths (a newer release in the same major version, the next major version) are indeed supported.
That being said, while staying at a half-upgraded state for a long time is possible, it is not recommended. It also means that whatever new features or improved algorithms were introduced in the new version, the new nodes will need to avoid them until the full cluster is upgraded.
OSS 4.1.7 is a pretty old OSS release from Oct 2020. The assumption that you can add another DC running OSS 5.0 (latest OSS release from 10 days ago) to the existing cluster, is a bit of a risky one.
The supported upgrade path (QA tested) is from OSS 4.6 to 5.0. You can read more about the upgrade path here: https://docs.scylladb.com/upgrade/upgrade-opensource/
The tested upgrade route is via minor versions 4.1 --> 4.2 --> 4.3 --> 4.4 --> 4.5 --> 4.6 --> 5.0, jumping multiple minor version should work, but we can't say that it was tested.
I'm trying to determine the correct upgrade process for a 3 node cluster running 9.x sitting behind a load balancer.
Looking at the cluster configuration we find :
Node 1 has forests and is the boot node ( likely the d-node ) and 300GB free.
Node 2 has no forests and has 700 GB free disk space ( likely an e-node ).
Node 3 has no forests and has 700 GB free disk space ( likely an e-node ).
Can I confirm that the correct cluster upgrade sequence from 9.x to 10.x Server is :
Shutdown node 3,2,1 ( in that order )
Deinstall converters then deinstall Server 9.x from Node 1
Deinstall converters then deinstall Server 9.x from Node 2 and Node 3
Install Server 10.x then install converters on Node 1 & restart Node 1
Install Server 10.x then install converters on Node 2 & 3 and restart both nodes.
Have I missed anything?
Is there anything I can look for on each node that tells me specifically that a node is an e-node or d-node?
Thanks.
Unless you are attempting to perform a rolling-upgrade (probably not necessary for a small 3-node cluster that only has one D-node), it is easiest to just shut them all down, uninstall the converters and MarkLogic, install the new version of server and converters, then start them all up and complete the upgrade in the Admin UI.
https://docs.marklogic.com/guide/installation/procedures#id_77358
Stop MarkLogic Server (as described in step 1 of Removing MarkLogic).
Uninstall the old MarkLogic 9 release (as described in Removing MarkLogic).
If you want to uninstall MarkLogic 9.0-4 or later, and if the converters package was previously installed with it, you will have to perform a two-step uninstall: first uninstall MarkLogic Converters and then uninstall MarkLogic Server. For more detail, see MarkLogic Converters Installation Changes Starting at Release 9.0-4 and Removing MarkLogic.
Install the new MarkLogic 10 release (as described in Installing MarkLogic).
If you want to install MarkLogic 9.0-4 or later, and you plan to use the converters package with it, you will have to perform a two-step installation: first install MarkLogic Server and then install MarkLogic Converters. For more detail, see MarkLogic Converters Installation Changes Starting at Release 9.0-4 and Installing MarkLogic.
Start MarkLogic Server (as described in Starting MarkLogic Server).
Open the Admin Interface in a browser (http://localhost:8001/).
When the Admin Interface prompts you to upgrade the databases and the configuration files, click the button to confirm the upgrade.
I've installed redis-server using apt-get install redis-server and everything went fine.
Right now I'm trying to configure it in a Cluster mode. The problem is that in the tutorial supplied here http://redis.io/topics/cluster-tutorial they use a script called redis-trib.rb which I can't find it in my system.
Can you please tell me how can I configure my Redis to run in Cluster mode without that script ?
I would like to have a setup with two masters, each on a different machine.
Thank you very much.
Had same problem with reredis-trib.rb
This tutorial explains how to create Redis Cluster using only Redis commands: Configuring and Running Redis Cluster on Linux
You need Redis 3.0.0 beta to run Cluster. You'll not find it in a Linux distribution, since they all have copy of the stable server (fortunately!). Redis 3.0.0 will go out as a stable release the next week. You can find the source code of the stable release here: http://redis.io/download.
There is now a tutorial for Ubuntu at https://www.digitalocean.com/community/tutorials/how-to-configure-a-redis-cluster-on-ubuntu-14-04 which includes installation of a PPA to supply 3.0.x. This tutorial is only for two nodes and does not reference redis-trib.rb ...
I have the following already setup (in order)
A Hadoop 2.5.x - 3 nodes cluster.
Apache Ambari Version 1.7.0
So my question is, Is there anyway I can integrate/configure Amabri to monitor my already setup Hadoop cluster ?
Any suggestions/views are much appreciate.
Currently it seems like that's not possible, there is open JIRA ticket on this issue:
https://issues.apache.org/jira/browse/AMBARI-2852
3.9.71 is the latest BSD versionned version of ServiceStack.Redis.
But Redis goes on with BSD licenses.
There are merge required changes for redis (2.6.X -> 2.8.X)
https://raw.github.com/antirez/redis/2.8/00-RELEASENOTES
So the question: Could we upgrade Redis to 2.8.X without upgrading ServiceStack.Redis?
(Production environment)
Thanks