I am running 3 node apache cassandra cluster as docker container holding timeseries data with 45 days TTL.
I am planning to upgrade the current cassandra version 2.2.5 to cassandra 3.11.4 release. Following steps are identified for the upgrade -
Backup existing data
Flush one of the cassandra node
bin/nodetool -h cassandra1 -u ca_itoa -pw ca_itoa drain
Stop the cassandra1 node
Start the new cassandra 3.11.4 container
Upgrade the SSTable
bin/nodetool -u ca_itoa -pw ca_itoa upgradesstables
Check the node status. Repeat the process for the rest of the nodes
I have few questions about the upgrade process -
Are the steps correct?
Is it manodatory to run upgradesstables command. It is time consuming, and I want to see if I can avoid. The data has TTL set. Will the cassandra continue writing in new SSTable format whereas the old SSTable data get cleaned-up on expiring? Assumption is that, after 45 days, all SSTable would be in new shiny format.
Just some additional thoughts:
For Step #6, you actually don't have to run upgradesstables right away. In fact, if you're upgrading a production system, it's probably better that you don't until the application team verifies that they can connect ok. Remember, older versions of the driver which work in 2.2 may not work with 3.11.4.
To this end, I would wait until the entire cluster is running on the new version before running upgradesstables on each node.
Is it manodatory to run upgradesstables command?
As each Cassandra version is capable of reading its own SSTable format as well as the prior major version, I guess it's not mandatory. But it's definitely something that you should want to do. Especially when upgrading to 3.x.
Cassandra 3 contains a significant upgrade to the storage engine, which results is a much smaller disk footprint. One cluster I upgraded saw a 90% reduction in disk needs.
Plus, you'd be incurring additional latency when reading records which may be spread across the old SSTable files as well as the new. Reads for records across multiple files are bad enough as it is. But now you'd be forcing Cassandra to read and collate results from two formats.
So while I wouldn't say it's "mandatory," I'd definitely say it qualifies as a "good idea."
Yes, you need to run nodetool sstableupgrade on each node after cassandra upgrade as you are upgrading from 2.2.x to 3.11.4. sstable file format and ext also will change. You may run this process on background and it will not create any issue. please refer below links for more details https://blog.thethings.io/upgrading-apache-cassandra-cluster/
Related
I want to upgrade cassandra 2.1.19 cluster to 3.11.1 without downtime.
Will 3.11.1 nodes work together with 2.1.19 nodes at the same time?
Key point will be how you connect to your cluster. You will need to try out on test systems if everything works from your application side doing the switch.
I recommend a two stop process in this case, migrate from 2.1.19 to 3.0.x - one node at atime.
For every node do the following (i said you need to test before before going to production right?):
nodetool drain - wait for finish
stop cassandra
backup your configs, the old one wont work out of the box
remove the cassandra package / tarball
read about the java and other cassandra 3.x requirements and ensure you met them
add the repo and install the 3.0.x package or tar ball
some packages start the node immediately - you may have to stop them again
make up the new config files (diff or something will be your friend, read the docs about the new options), one time only you should be able to resuse if on all the other nodes
start cassandra (did I say test this on a test system?) and wait until the node has joined the ring again nodetool status
upgrade your sstables with nodetool upgradesstables - almost always needed, dont skip this even if "something" works right now
this upgrade tends to be really slow - it's just a single thread running rewriting all your data, so i/o will be a factor here
all up and running -> go ahead to the next node and repeat
After that - upgrade 3.0.x to 3.11.x in the same fashion, add the new repo, configure for 3.11.x as for 3.0.x above and so on. But this time you can skip upgrading sstables as the format stays the same (but it wont harm if you do so).
Did I mention to do this on testing system first? One thing that will happen and may break things - older native protocols will be gone as well as rpc/ thrift.
Hope I didn't miss something ;)
We are upgrading from DSE 4.5 to DSE 4.8.9 in 10 node PRODUCTION cluster.
We have daily batch jobs running in our application which bulk load the data in the cluster , some jobs TRUNCATE the tables and load fresh data and some loader jobs which continuously insert the data.
Consider these scenarios :
Case 1 :
Let say my one node has DSE 4.8 installed but upgradesstables is running .
All nodes are online at this moment and 2 different schema exist (9 nodes on dse4.5 and 1 node on dse4.8.9).
In this case , will TRUNCATE work ?
Case 2:
One of my nodes is fully upgraded to DSE 4.8 , which makes my cluster to be in partially upgraded state, all nodes online,2 schema exists (9 nodes on DSE 4.5 and 1 node on DSE 4.8).
Will TRUNCATE work in this case ?
Please suggest.
Thanks!
Its not recommended to issue a TRUNCATE command during an upgrade, this is one of the limitations outlined here
To quote the link:
Do not enable new features.
Do not run nodetool repair.
Do not issue these types of CQL queries during a rolling restart: DDL
and TRUNCATE.
During upgrades, the nodes on different versions might show a schema
disagreement.
Failure to upgrade SSTables when required results in a significant
performance impact and increased disk usage. Upgrading is not complete
until the SSTables are upgraded.
It should be a practice to upgrade the binaries first on all the nodes so that we have one schema across the cluster .
Avoid the use of TRUNCATE till all nodes have completed running "upgradesstables".
Comment given by markc is also to be noted :
Do not enable new features.
Do not run nodetool repair.
Do not issue these types of CQL queries during a rolling restart: DDL and TRUNCATE.
During upgrades, the nodes on different versions might show a schema disagreement.
Failure to upgrade SSTables when required results in a significant performance impact and increased disk usage. Upgrading is not complete until the SSTables are upgraded.
I just have one reason to restart cluster below :
All the nodes have the same hardware configuration
1. When i update file cassandra.yaml
Are there other reasons ?
The thing you are asking for is Rolling Restart a cassandra cluster. There are so many reason to restart a cassandra cluster. I'm just mentioning some below-
when you update any value in cassandra.yaml. (As you mentioned above)
When your nodetool got stucked somehow. such as- you gave command nodetool repair and cancelled the command but it got stucked behind, then you won't be able to give another nodetool repair command.
When you are adding a new node to cluster and you got stream_failed due to nproc limit. That time your running cluster nodes could be down to this issue and going to hold the status.
When you don't want to use sstableloader and you need to restore your data from snapshots. That time you need to provide your snapshots to the data_directory on each node and rolling restart.
When you are about to upgrade your cassandra_version.
For example when you upgrading Cassandra version.
I've searched for previous versions of this question, but none seem to fit my case. I have an existing Cassandra cluster running 2.0.x. I've been allocated new VMs, so I do NOT want to upgrade my existing Cassandra nodes - rather I want to migrate to a) new VMs and b) a more current version of Cassandra.
I know for in-place upgrades, I would upgrade to the latest 2.0.x, then to the latest 2.1.x. AFAIK, there's no SSTable inconsistency here. If I go this route via addition of new nodes, I assume I would follow the datastax instructions for adding new nodes/decommissioning old nodes?
Given the above, is it possible to move from 2.0.x to 3.0.x? I know the SSTable format is different; however, if I'm adding new nodes (rather than re-using SSTables on disk), does this matter?
It seems to me that #2 has to work - otherwise, it implies that any upgrade requiring SSTable upgrades would require all nodes to be taken offline simultaneously; otherwise, there would be mixed 2.x.x and 3.0.x versions running in the same cluster at some point.
Am I completely wrong? Does anyone have any experience doing this?
Yes, it is possible to migrate data to a different environment (the new vm's with the updated Cassandra using sstableloader, but you will need C* 3.0.5 and above, as that version added support to upload sstables from previous versions.
Once that the process is completed it is recommended to execute nodetool upgradesstables to ensure that there are no incompatibilities on the data, and a nodetool cleanup.
Regarding your comment ... it implies that any upgrade requiring SSTable upgrades would require all nodes to be taken offline simultaneously;... is not true; doing the upgrade one node at a time will create a mixed cluster with nodes with the two versions as you mentioned, which is not optimal, but will allow you to avoid any downtime in production. (Note that the impact of this operation will depend on the consistency level used in your application.)
Don't worry about the migration. You can simply migrate your Cassandra 2.0.X cluster to Cassandra 3.0.X. But its better if you migrate your cluster Cassandra 2.0.X to latest Cassandra 2.X.X then Cassandra 3.0.X. You need to follow some steps-
Backup data
Uninstall present version
Install the version you want to upgrade
Restore data
As you are doing migration, you need to be careful about your data always. For the data backup and restore you can follow two ways-
Creating snapshots of your sstables and then after installing the new version of cassandra, placing the files to the data location and run sstableloader.
Backup your schema's to a .cql file and copy all the tables to .csv and then after installing the new version of cassandra, source your schema from .cql and copy all the tables from every single .csv file.
If you are fully convinced how you will complete the migration then you can write a bash script to complete the backup and restore steps.
My current cassandra version is 1.1.2, it is implemented with a single node cluster, i would like to upgrade it 1.2.6 with multiple nodes in the ring. is it a proper way to migrate it directly to 1.2.6 or i should follow version by version migration.
I found the upgrading steps from this link
http://fossies.org/linux/misc/apache-cassandra-1.2.6-bin.tar.gz:a/apache-cassandra-1.2.6/NEWS.txt.
There are 9 other releases available between this two versions.
I migrate a two cluster nodes from 1.1.6 to 1.2.6 without problems and without doing version by version. Anyway, you should take a closer look into:
http://www.datastax.com/documentation/cassandra/1.2/index.html?pagename=docs&version=1.2&file=index#upgrade/upgradeC_c.html#concept_ds_smb_nyr_ck
Because there are a lot of new features from version 1.2 like the partioners maybe you need to change some configurations for your cluster.
You may directly hop on to C1.2.6.
We migrated our 4-node cluster from C1.0.9 to C1.2.8 recently without any issues. This was a rolling upgrade i.e. upgrade one node at a time and after each upgrade of a node, allow the cluster to stabilize (depends upon the traffic during upgrade)
These are the steps that we followed:
Perform below on each node,
Run Disablegossip and disablethrift, such that this node is seen as DOWN by other nodes.
flush/drain the memtables, run compaction to merge SSTables
take snapshot and enable incremental backups
This stops all the other nodes/clients from writing to this node and since memtables are flushed to disk, startup times are fast as it need not walk-through commit logs.
stop Cassandra (though this node is down, cluster is available for write/read, so zero downtime)
upgrade sstables to new storage format using sstableupgrade
install/untar Cassandra 1.2.8 on the new locations
move upgraded sstables to appropriate location
merge Cassandra.yaml from previous version and current version by a manual diff (need to detail out difference)
start Cassandra
watch the startup messages to ensure the node comes up without difficulty and is shown in the ring with mixed 1.0.x/1.2.x