The k8ssandra requirement is a major blocker.
Sean R. Durity
INTERNAL USE
From: Christopher Bradford
Sent: Tuesday, February 27, 2024 9:49 PM
To: user@cassandra.apache.org
Cc: Christopher Bradford
Subject: [EXTERNAL] Re: Check out new features in K8ssandra and Mission Control
Hey Jon, * What
Cost of availability is a fair question at some level of the discussion. In my
experience, high availability is one of the top 2 or 3 reasons why Cassandra is
chosen as the data solution. So, if I am given a Cassandra use case to build
out, I would normally assume high availability is needed, ev
For a variety of reasons, we have clusters with 5 TB of disk per host as a
“standard.” In our larger data clusters, it does take longer to add/remove
nodes or do things like upgradesstables after an upgrade. These nodes have 3+TB
of actual data on the drive. But, we were able to shrink the node
I would expect single digit ms latency on reads and writes. However, we have
not done any performance testing on Apache Cassandra 4.x.
Sean R. Durity
INTERNAL USE
From: Shaurya Gupta
Sent: Friday, August 11, 2023 1:16 AM
To: user@cassandra.apache.org
Subject: [EXTERNAL] Re: Cassandra p95 laten
We also parse the output from nodetool info and nodetool status and (to a
lesser degree) nodetool netstats. We have basically made info and status more
operator-friendly in a multi-cluster environment. (And we added a useable
return value to our info command that we can use to evaluate the node’
I run clean-up in parallel, not serially, since it is a node-only kind of
operation. And I only run in the impacted DC. With only 300 GB on a node,
clean-up should not take very long. Check your compactionthroughput.
I ran clean-up in parallel on 53 nodes with over 3 TB of data each. It took
li
n Thu, Feb 16, 2023 at 9:43 AM Marc Hoppins
mailto:marc.hopp...@eset.com>> wrote:
Compaction_throughtput_per_mb is 0 in cassandra.yaml. Is setting it in nodetool
going to provide any increase?
From: Durity, Sean R via user
mailto:user@cassandra.apache.org>>
Sent: Thursday, February 16,
Clean-up is constrained/throttled by compactionthroughput. If your system can
handle it, you can increase that throughput (nodetool setcompactionthroughput)
for the clean-up in order to reduce the total time.
It is a node-isolated operation, not cluster-involved. I often run clean up on
all nod
In most cases, I would delete the corrupt commit log file and restart. Then run
repairs on that node. I have seen cases where multiple files are corrupted and
it is easier to remove all commit log files to get the node restarted.
Sean R. Durity
From: Joe Obernberger
Sent: Friday, February 3, 20
For physical hardware when disks fail, I do a removenode, wait for the drive to
be replaced, reinstall Cassandra, and then bootstrap the node back in (and run
clean-up across the DC).
All of our disks are presented as one file system for data, which is not what
the original question was asking.
: [EXTERNAL] Re: Best compaction strategy for rarely used data
On 2022-12-29 21: 54, Durity, Sean R via user wrote: > At some point you will
end up with large sstables (like 1 TB) that won’t > compact because there are
not 4 similar-sized ones able to be compacted Yes, that's exa
If there isn’t a TTL and timestamp on the data, I’m not sure the benefits of
TWCS for this use case. I would stick with size-tiered. At some point you will
end up with large sstables (like 1 TB) that won’t compact because there are not
4 similar-sized ones able to be compacted (assuming default
I have seen this when there is a tab character in the yaml file. Yaml is (too)
picky on these things.
Sean R. Durity
DB Solutions
Staff Systems Engineer – Cassandra
From: Amit Patel via user
Sent: Thursday, December 8, 2022 11:38 AM
To: Arvydas Jonusonis ; user@cassandra.apache.org
Subject: [EX
Does it need to be strictly Apache Cassandra? Or is something built on/working
with DataStax Enterprise allowed? I would think if it doesn’t depend on
DSE-only technology, it could still apply to a general Cassandra audience.
Sean R. Durity
From: Patrick McFadin
Sent: Tuesday, November 29, 20
>From the subject, this looks like a client-side timeout (thrown by the
>driver). I have seen situations where the client/driver timeout of 2 seconds
>is a shorter timeout than on the server side (10 seconds). So, the server
>doesn’t really note any problem. Unless this is a very remote client a
Aggregate queries (like count(*) ) are fine *within* a reasonably sized
partition (under 100 MB in size). However, Cassandra is not the right tool if
you want to do aggregate queries *across* partitions (unless you break up the
work with something like Spark). Choosing the right partition key a
In my experience C* is not cheaper storage than HDFS. If that is the goal, it
may be painful.
Each Cassandra DC has at least one full copy of the data set. For production
data that I care about (that my app teams care about), we use RF=3 in each
Cassandra DC. And I only use 1 Cassandra rack per
I'm not afraid to admit that I LOVE this feature. Exactly what a data engine
should be able to do - stop bad behavior.
Sean R. Durity
From: Aaron Ploetz
Sent: Thursday, June 23, 2022 3:22 PM
To: user@cassandra.apache.org
Subject: [EXTERNAL] Re: Guardrails in Cassandra 4.1 Alpha
Ahh...yes, my d
It can work to use host names. We have done it for temporary clusters where
there is at least a theoretical possibility of an ip address change. I don't
know all the trade-offs of using host names, since we don't do that for
production.
Sean R. Durity
INTERNAL USE
-Original Message-
I have run clusters with different disk size nodes by using different number of
num_tokens. I used the basic math of just increasing the num_tokens by the same
percentage as change in disk size. (So, if my "normal" node was 8 tokens, one
with double the disk space would be 16.)
One thing to wat
02/06/2022 14:46, Durity, Sean R wrote:
I agree with Marc. We use the cassandra-topology.properties file (and
PropertyFileSnitch) for our deployments. Having a file different on every node
has never made sense to me. There would still have to be some master file
somewhere from which to generate
I agree with Marc. We use the cassandra-topology.properties file (and
PropertyFileSnitch) for our deployments. Having a file different on every node
has never made sense to me. There would still have to be some master file
somewhere from which to generate that individual node file. There is the
A select with no where clause is not a good access pattern for Cassandra,
regardless of driver version. It will not scale for large data sets or a large
number of nodes.
Ideally you want to select from a single partition for each query. So,
depending on the size of the rows, one answer may be t
If the number of rows is known and bounded and would be under 100 MB in size, I
would suggest adding an artificial partition key so that all rows are in one
partition. I recommend this technique for something like an application
settings table that is retrieved infrequently (like on app start-up
I have used my own bash scripts with ssh connections to the nodes to automate
everything from upgrades, node down monitoring, metrics or log collection, and
rolling restarts. We are moving toward ansible (our infrastructure team is
standardizing on its use). Rolling restart isn’t too bad in ansi
Has anyone been able to add Apache Cassandra 4.x nodes to a new DC within a DSE
6+ cluster (or vice versa) in order to migrate from one to the other with no
downtime? I was able to do this prior to DSE 6/Cassandra 4.0, but that was
before the internals rewrite (and different sstable format?) of
In my experience, the 50% overhead for compaction/upgrade is for the worst case
scenario systems – where the data is primarily one table and uses size-tiered
compaction. (I have one of those.) What I really look at is if there is enough
space to execute upgradesstables on the largest sstable. Gr
We have apps like this, also. For straight Cassandra, I think it is just the
nature of how it works. DataStax provides some interesting solutions in
different directions: BigNode (for handling 10-20 TB nodes) or Astra
(cloud-based/container-driven solution that DOES separate read, write, and
st
For memory-sake, you do not want “too many” tables in a single cluster (~200 is
a reasonable rule of thumb). But I don’t see a major concern with a few very
large tables in the same cluster. The client side, at least in Java, could get
large (memory-wise) holding a Cluster object for multiple cl
Each table in the cluster will have a memtable. This is why you do not want to
fracture the memory into 900+ slices. The rule of thumb I have followed is to
stay in the low hundreds (maybe 200) tables for the whole cluster. I would be
requiring the hard refactoring (or moving tables to different
Yes, use the time-bucketing approach and choose a bucket-size (included in the
partition key) that is granular enough to keep partitions to about 100 MB in
size. (Unbounded partitions WILL destroy your cluster.) If your queries *need*
to retrieve all user activity over a certain period, then, ye
There is not enough 4.0 chatter here. What feature or fix of the 4.0 release is
most important for your use case(s)/environment? What is working well so far?
What needs more work? Is there anything that needs more explanation?
[cid:image001.png@01D7431D.8C1332E0]
Sean Durity
Staff Systems Engine
Try adding this into the SSL section of your cqlshrc file:
version = SSLv23
Sean Durity
From: Maxim Parkachov
Sent: Friday, April 30, 2021 8:57 AM
To: user@cassandra.apache.org; d...@cassandra.apache.org
Subject: [EXTERNAL] Cassandra 3.11 cqlsh doesn't work with latest JDK
Hi everyone,
I have
DataStax Enterprise has a new-ish feature set called Big Node that is supposed
to help with using much denser nodes. We are going to be doing some testing
with that for a similar use case with ever-growing disk needs, but no real
increase in read or write volume. At some point it may become avai
I have a cluster (almost 200 nodes) with a variety of disk sizes and use
different numbers of tokens so that the machines can use the disk they have. It
is a very handy feature! While I agree that a node with larger disk may handle
more requests, that may not be enough to impact CPU or memory. I
+1 for data modeling. If an admin can spend the day helping app teams get the
model right BEFORE hitting production, those are the best days (and prevent the
bad days of trying to engineer around a bad model/migrate data to new
tables/etc)
I also find good value in understanding the availabilit
Sometimes time bucketing can be used to create manageable partition sizes. How
much data is attached to a day, week, or minute? Could you use a partition and
clustering key like: ((source, time_bucket), timestamp)?
Then your application logic can iterate through time buckets to pull out the
dat
Are there specific queries that are slow? Partition-key queries should have
read latencies in the single digits of ms (or faster). If that is not what you
are seeing, I would first review the data model and queries to make sure that
the data is modeled properly for Cassandra. Without metrics, I
I would be looking at the queries in the application to see if there are any
cross-partition queries (ALLOW FILTERING or IN clauses across partitions). This
looks like queries that work fine with small scale, but are hitting timeouts
when the data size has increased.
Also see if anyone has ad-ho
machines.
What would you do to build a 1PByte configuration? I know there are a lot of -
it depends - on that question, but say it was a write heavy, light read setup.
Thank you!
-Joe
On 1/20/2021 10:06 AM, Durity, Sean R wrote:
Yakir is correct. While it is feasible to have large disk
Yakir is correct. While it is feasible to have large disk nodes, the practical
aspect of managing them is an issue. With the current technology, I do not
build nodes with more than about 3.5 TB of disk available. I prefer 1-2 TB, but
costs/number of nodes can change the considerations.
Putting
This may not answer all your questions, but maybe it will help move you further
along:
- you could copy the data (not system) folders *IF* the clusters match in
topology. This would include the clusters having the same token range
assignment(s). And you would have to copy the folders from one or
Tombstone_failure_threshold is only for reads. If the tombstones are in
different partitions, and you aren’t doing cross-partition reads, you shouldn’t
need to adjust that value.
If disk space recovery is the goal, it depends on how available you need the
data to be. The faster way is probably
slow reads due to compaction / IO / gc /
slow disks / bad memory / etc).
On Wed, Nov 11, 2020 at 6:36 AM Durity, Sean R
mailto:sean_r_dur...@homedepot.com>> wrote:
Doesn’t the read get sent to all nodes that own the data in parallel (from the
coordinator)? And the first one that is able to r
Doesn’t the read get sent to all nodes that own the data in parallel (from the
coordinator)? And the first one that is able to respond wins (for LOCAL_ONE).
That was my understanding.
Sean Durity
From: Jeff Jirsa
Sent: Wednesday, November 11, 2020 9:24 AM
To: user@cassandra.apache.org
Subject:
Auth
Sent: Tuesday, November 10, 2020 11:50 AM
To: user@cassandra.apache.org
Subject: [EXTERNAL] Re: Last stored value metadata table
Hi,
On Tue, Nov 10, 2020 at 5:29 PM Durity, Sean R
mailto:sean_r_dur...@homedepot.com>> wrote:
Updates do not create tombstones. Deletes create tombstone
Hi,
On Tue, Nov 10, 2020 at 3:18 PM Durity, Sean R
mailto:sean_r_dur...@homedepot.com>> wrote:
My answer would depend on how many “names” you expect. If it is a relatively
small and constrained list (under a few hundred thousand), I would start with
something like:
At the moment, the
My answer would depend on how many “names” you expect. If it is a relatively
small and constrained list (under a few hundred thousand), I would start with
something like:
Create table last_values (
arbitrary_partition text, -- use an app name or something static to define the
partition
name tex
I’m a little late on this one, but I would choose approach 1. It is much more
comprehensible to anyone who comes afterwards. And it should easily scale in
Cassandra to whatever volume you have. I think I would call the table
recent_users to make it very clear the purpose of the table. It is also
I would use dsbulk to unload and load. Then the schemas don’t really matter.
You define which fields in the resulting file are loaded into which columns.
You also won’t have the limitations and slowness of COPY TO/FROM.
Sean Durity
From: Mitch Gitman
Sent: Friday, July 24, 2020 2:22 PM
To: us
I’m curious – is the scaling needed for the amount of data, the amount of user
connections, throughput or what? I have a 200ish cluster, but it is primarily a
disk space issue. When I can have (and administer) nodes with large disks, the
cluster size will shrink.
Sean Durity
From: Isaac Reath
I plan downtime for changes to security settings like this. I could not come up
with a way to not have degraded access or inconsistent data or something else
bad. The foundational issue is that unencrypted nodes cannot communicate with
encrypted ones.
I depend on Cassandra’s high availability f
the nodes with 3.11.3.
On Wednesday, June 24, 2020, Durity, Sean R
mailto:sean_r_dur...@homedepot.com>> wrote:
Streaming operations (repair/bootstrap) with different file versions is usually
a problem. Running a mixed version cluster is fine – for the time you are doing
the upgrade. I would no
Streaming operations (repair/bootstrap) with different file versions is usually
a problem. Running a mixed version cluster is fine – for the time you are doing
the upgrade. I would not stay on mixed versions for any longer than that. It
takes more time, but I separate out the admin tasks so that
Sent: Tuesday, June 2, 2020 10:48 AM
To: user@cassandra.apache.org
Subject: [EXTERNAL] Re: Cassandra Bootstrap Sequence
3000 tables
On Tuesday, June 2, 2020, Durity, Sean R
mailto:sean_r_dur...@homedepot.com>> wrote:
How many total tables in the cluster?
Sean Durity
From: Jai Bheems
To flesh this out a bit, I set roles_validity_in_ms and
permissions_validity_in_ms to 360 (10 minutes). The default of 2000 is far
too often for my use cases. Usually I set the RF for system_auth to 3 per DC.
On a larger, busier cluster I have set it to 6 per DC. NOTE: if you set the
validi
How many total tables in the cluster?
Sean Durity
From: Jai Bheemsen Rao Dhanwada
Sent: Monday, June 1, 2020 8:36 PM
To: user@cassandra.apache.org
Subject: [EXTERNAL] Re: Cassandra Bootstrap Sequence
Thanks Erick,
I see below tasks are being run mostly. I didn't quite understand what exactly
I thought this might be a single-time use case request. I think my first
approach would be to use something like dsbulk to unload the data and then
reload it into a table designed for the query you want to do (as long as you
have adequate disk space). I think like a DBA/admin first. Dsbulk creat
I agree – do not aim for a mixed version as normal. Mixed versions are fine
during an upgrade process, but the goal is to complete the upgrade as soon as
possible.
As for other parts of your plan, the Kafka Connector is a “sink-only,” which
means that it can only insert into Cassandra. It doesn
Oh, I see it was clock drift in this case. Glad you found that out.
Sean Durity
From: Durity, Sean R
Sent: Tuesday, March 24, 2020 2:10 PM
To: user@cassandra.apache.org
Subject: [EXTERNAL] RE: Table not updating
I’m wondering about nulls. They are written as tombstones. So, it is an
I’m wondering about nulls. They are written as tombstones. So, it is an
interesting question for a prepared statement where you are not binding all the
variables. The driver or framework might be doing something you don’t expect.
Sean Durity
From: Sebastian Estevez
Sent: Monday, March 23, 2020
I agree. Cassandra already hashes the partition key to a numeric token.
Sean Durity
From: Jon Haddad
Sent: Friday, March 6, 2020 9:29 AM
To: user@cassandra.apache.org
Subject: [EXTERNAL] Re: Performance of Data Types used for Primary keys
It's not going to matter at all.
On Fri, Mar 6, 2020, 2
I agree – a back out becomes practically very challenging after the second node
is upgraded, because the new data is written in the new disk format. To satisfy
the “you must have a backout” rules, I just say that after node 1, I could stop
that node, wipe the data, downgrade the binaries, and re
Batches are for atomicity, not performance.
I would do single deletes with a prepared statement. An IN clause causes extra
work for the coordinator because multiple partitions are being impacted. So,
the coordinator has to coordinate all nodes involved in those writes (up to the
whole cluster).
I would consider building a lookup table instead. Something like:
Create table new_lookup (
new-lookup-partition text,
existing-key text
PRIMARY KEY (new-lookup-partition)
)
For me, these are easier to understand and reason through for Cassandra
performance and availability. I would use
I would also push for something besides a full refresh, if at all possible. It
feels like a waste of resources to me – and not predictably scalable.
Suggestions: use a queue to send writes to both systems. If the downstream
system doesn’t handle TTL, perhaps set an expiration date and a purge qu
+1 on nodetool drain. I added that to our upgrade automation and it really
helps with post-upgrade start-up time.
Sean Durity
From: Erick Ramirez
Sent: Wednesday, February 12, 2020 10:29 PM
To: user@cassandra.apache.org
Subject: Re: [EXTERNAL] Cassandra 3.11.X upgrades
Yes to the steps. The on
I will just add-on that I usually reserve security changes as the primary
exception where app downtime may be necessary with Cassandra. (DSE has some
Transitional tools that are useful, though.) Sometimes a short outage is
preferred over a longer, more-complicated attempt to keep the app up. And
Ah - I should have looked it up! Thank you for fixing my mistake.
Sean Durity
-Original Message-
From: Michael Shuler
Sent: Wednesday, February 12, 2020 3:17 PM
To: user@cassandra.apache.org
Subject: Re: [EXTERNAL] Cassandra 3.11.X upgrades
On 2/12/20 12:58 PM, Durity, Sean R wrote
est,
Sergio
On Wed, Feb 12, 2020, 10:58 AM Durity, Sean R
mailto:sean_r_dur...@homedepot.com>> wrote:
Check the readme.txt for any upgrade notes, but the basic procedure is to:
* Verify that nodetool upgradesstables has completed successfully on all
nodes from any previous upgrade
*
Check the readme.txt for any upgrade notes, but the basic procedure is to:
* Verify that nodetool upgradesstables has completed successfully on all
nodes from any previous upgrade
* Turn off repairs and any other streaming operations (add/remove nodes)
* Stop an un-upgraded node (seed
This looks like an error between your client and the cluster. Is the other ip
address your client app? I have typically seen this when there are network
issues between the client and the cluster. Cassandra driver connections are
typically very long-lived. If something like a switch or firewall t
_session (
userid bigint,
session_usr text,
last_access_time timestamp,
login_time timestamp,
status int,
PRIMARY KEY (userid, session_usr)
) WITH CLUSTERING ORDER BY (session_usr ASC)
On Thu, Feb 6, 2020 at 2:09 PM Durity, Sean R
mailto:sean_r_dur...@homedepot.com>> wrote:
Do yo
LUSTERING ORDER BY (session_usr ASC)
On Thu, Feb 6, 2020 at 2:09 PM Durity, Sean R
mailto:sean_r_dur...@homedepot.com>> wrote:
Do you only need the current count or do you want to keep the historical counts
also? By active users, does that mean some kind of user that the application
t
Do you only need the current count or do you want to keep the historical counts
also? By active users, does that mean some kind of user that the application
tracks (as opposed to the Cassandra user connected to the cluster)?
I would consider a table like this for tracking active users through ti
)
* saved_caches (usually located in /var/lib/cassandra/saved_caches)
Cheers,
Anthony
On Fri, 31 Jan 2020 at 03:05, Durity, Sean R
mailto:sean_r_dur...@homedepot.com>> wrote:
Your procedure won’t work very well. On the first node, if you switched to 4,
you would end up with only a tiny
Your procedure won’t work very well. On the first node, if you switched to 4,
you would end up with only a tiny fraction of the data (because the other nodes
would still be at 256). I updated a large cluster (over 150 nodes – 2 DCs) to
smaller number of vnodes. The basic outline was this:
*
I would suggest to be aware of potential data size expansion. If you load (for
example) three copies of the data into a new cluster (because the RF of the
origin cluster is 3), it will also get written to the RF of the new cluster (3
more times). So, you could see data expansion of 9x the origin
sstablekeys (in the tools directory?) can extract the actual keys from your
sstables. You have to run it on each node and then combine and de-dupe the
final results, but I have used this technique with a query generator to extract
data more efficiently.
Sean Durity
From: Chris Splinter
Sent:
is not an issue - I guess DSBulk or
SSTableloader would be an option? 2. If merge is an issue - I am guessing
without app code change - this wont be possible ,right?
Thanks & Regards,
Ankit Gadhiya
On Fri, Jan 17, 2020 at 9:40 AM Durity, Sean R
mailto:sean_r_dur...@homedepot.com>> wro
A couple things to consider:
* A separation of apps into their own clusters is typically a better model
to avoid later entanglements
* Dsbulk (1.4.1) is now available for only open source clusters. It is a
great tool for unloading/loading
* What data problem are you trying to solve w
I use a script that calls nodetool info. If nodetool info returns an error
(instance isn’t up, on the way up, etc.) then I return that same error code
(and I know the node is NOT OK). If nodetool info succeeds, I then parse the
output for each protocol to be up. A node can be up, but have gossip
100,000 rows is pretty small. Import your data to your cluster, do a nodetool
flush on each node, then you can see how much disk space is actually used.
There are different compression tools available to you when you create the
table. It also matters if the rows are in separate partitions or you
Any read-only file systems? Have you tried to start from the command line
(instead of a service)? Sometimes that will give a more helpful error when
start-up can’t complete.
If your error is literally what you included, it looks like the executable
can’t find the cassandra.yaml file.
I will ag
The size of the data matters here. Copy to/from is ok if the data is a few
million rows per table, but not billions. It is also relatively slow (but with
small data or a decent outage window, it could be fine). If the data is large
and the outage time matters, you may need custom code to read fr
All my upgrades are without downtime for the application. Yes, do the binary
upgrade one node at a time. Then run upgradesstables on as many nodes as your
app load can handle (maybe you can point the app to a different DC, while
another DC is doing upgradesstables). Upgradesstables doesn’t cause
I’m not sure this is the fully correct question to ask. The size of the data
will matter. The importance of high availability matters. Performance can be
tuned by taking advantage of Cassandra’s design strengths. In general, you
should not be doing queries with a where clause on non-key columns.
There is definitely a resource risk to having thousands of open connections to
each node. Some of the drivers have (had?) less than optimal default settings,
like acquiring 50 connections per Cassandra node. This is usually overkill. I
think 5-10/node is much more reasonable. It depends on your
Everything in Cassandra is an insert. So, an update and an insert are
functionally equivalent. An update doesn't go update the existing data on disk;
it is a new write of the columns involved. So, the difference in your scenario
is that with the "targeted" update, you are writing less of the col
+1 for removing complexity to be able to create (and maintain!) “reasoned”
systems!
Sean Durity – Staff Systems Engineer, Cassandra
From: Reid Pinchback
Sent: Thursday, October 24, 2019 10:28 AM
To: user@cassandra.apache.org
Subject: [EXTERNAL] Re: Cassandra Rack - Datacenter Load Balancing re
Beneficial to whom? The apps, the admins, the developers?
I suggest that app teams have separate clusters per application. This prevents
the noisy neighbor problem, isolates any security issues, and helps when it is
time for maintenance, upgrade, performance testing, etc. to not have to
coordin
I don’t disagree with Jon, who has all kinds of performance tuning experience.
But for ease of operation, we only use G1GC (on Java 8), because the tuning of
ParNew+CMS requires a high degree of knowledge and very repeatable testing
harnesses. It isn’t worth our time. As a previous writer mentio
Copy command tries to export all rows in the table, not just the ones on the
node. It will eventually timeout if the table is large. It is really built for
something under 5 million rows or so. Dsbulk (from DataStax) is great for this,
if you are a customer. Otherwise, you will probably need to
DataStax has a very fast bulk load tool - dsebulk. Not sure if it is available
for open source or not. In my experience so far, I am very impressed with it.
Sean Durity – Staff Systems Engineer, Cassandra
-Original Message-
From: p...@xvalheru.org
Sent: Saturday, August 3, 2019 6:06 A
l the upgrade is done and
then change it back once the upgrade is completed?
On Fri, Jul 26, 2019 at 11:42 AM Durity, Sean R
mailto:sean_r_dur...@homedepot.com>> wrote:
What you have seen is totally expected. You can’t stream between different
major versions of Cassandra. Get the upgrade do
What you have seen is totally expected. You can’t stream between different
major versions of Cassandra. Get the upgrade done, then worry about any down
hardware. If you are using DCs, upgrade one DC at a time, so that there is an
available environment in case of any disasters.
My advice, though
This sounds like a bad query or large partition. If a large partition is
requested on multiple nodes (because of consistency level), it will pressure
all those replica nodes. Then, as the cluster tries to adjust the rest of the
load, the other nodes can get overwhelmed, too.
Look at cfstats to
The advice so far is exactly correct for an in-place kind of upgrade. The blog
post you mentioned is different. They decided to jump versions in Cassandra by
standing up a new cluster and using a dual-write/dual-read process for their
app. They also wrote code to read and interpret sstables in o
I’m not sure it is correct to say, “you cannot.” However, that is a more
complicated restore and more likely to lead to inconsistent data and take
longer to do. You are basically trying to start from a backup point and roll
everything forward and catch up to current.
Replacing/re-streaming is t
This may sound a bit harsh, but I teach my developers that if they are trying
to use ALLOW FILTERING – they are doing it wrong! We often choose Cassandra for
its high availability and scalability characteristics. We love no downtime.
ALLOW FILTERING is breaking the rules of availability and scal
1 - 100 of 203 matches
Mail list logo
