Using a hector Mutator I update some row over N sequential operation. Is there a guaranty, that changes happens in the order they where added to Mutator?
The simplest example, if I delete some row and then immediately recreate it. Could it not happen, that the deletion happens after inserting?
How cassandra cluster manages it, if two sequential requests are sent to different nodes? It is always possible there is few milliseconds difference between nodes...
Cassandra resolves conflicts using timestamps supplied by the client. In your example the 'recreate' of the row will have a higher timestamp than the row delete so it doesn't matter if somehow they got to the server in the wrong order.
One consequence of client supplied timestamps is that you either need to sync the clocks on your client machines or design your data model so that different clients don't conflict with each other.
Related
I am thinking about using Kafka connect to stream updates from Cassandra to a Kafka topic. The existing connector from StreamReactor seems to use a timestamp or uuidtimestamp to extract new changes since the last poll. The value of the timestamp is inserted using now() in the insert statement. The connector then saves the maximum time is received last time.
Since Cassandra is eventually consistent I am wondering what actually happens when doing repeated queries using a time range to get new changes. Is there not risk to miss rows inserted into Cassandra because it "arrived late" to the node queried when using WHERE create >= maxTimeFoundSoFar?
Yes it might happen that you have newer data in front of your "cursor" when you already went on with processing if you are using consistency level one for reading and writing, but even if you use higher consistency you might run into "problems" depending on the setup that you have. Basically there are a lot of things that can go wrong.
You can increase the chances of not doing this by using an old cassandra formula NUM_NODES_RESPONDING_TO_READ + NUM_NODES_RESPONDING_TO_WRITE > REPLICATION_FACTOR but since you are using now() from cassandra the node clocks might have millisecond offsets between them so you might even miss data if you have high frequency data. I know of some systems where people are actually using raspberry pi's with gps modules to keep the clock skew really tight :)
You would have to provide more about your use case but in reality yes you can totally skip some inserts if you are not "careful" but even then there is no 100% guarantee other then you process the data with some offset that would be enough for the new data to come in and settle.
Basically you would have to keep some moving time window in the past and then move it along plus making sure that you don't take into account anything newer than the let's say last minute. That way you are making sure the data is "settling".
I had some use cases where we processed sensory data that would came in with multiple days of delay. On some projects we simply ignored it on some the data was for reporting on the month level so we always processed the old data and added it to reporting database. i.e. we kept a time window 3 days back in history.
It just depends on your use case.
In the introduction course of Cassandra DataStax they say that all of the clocks of a Cassandra cluster nodes, have to be synchronized, in order to prevent READ queries to 'old' data.
If one or more nodes are down they can not get updates, but as soon as they back up again - they would update and there is no problem...
So, why Cassandra cluster need synchronized clocks between nodes?
In general it is always a good idea to keep your server clocks in sync, but a primary reason why clock sync is needed between nodes is because Cassandra uses a concept called 'Last Write Wins' to resolve conflicts and determine which mutation represents the most correct up-to date state of data. This is explained in Why cassandra doesn't need vector clocks.
Whenever you 'mutate' (write or delete) column(s) in cassandra a timestamp is assigned by the coordinator handling your request. That timestamp is written with the column value in a cell.
When a read request occurs, cassandra builds your results finding the mutations for your query criteria and when it sees multiple cells representing the same column it will pick the one with the most recent timestamp (The read path is more involved than this but that is all you need to know in this context).
Things start to become problematic when your nodes' clocks become out of sync. As I mentioned, the coordinator node handling your request assigns the timestamp. If you do multiple mutations to the same column and different coordinators are assigned, you can create some situations where writes that happened in the past are returned instead of the most recent one.
Here is a basic scenario that describes that:
Assume we have a 2 node cluster with nodes A and B. Lets assume an initial state where A is at time t10 and B is at time t5.
User executes DELETE C FROM tbl WHERE key=5. Node A coordinates the request and it is assigned timestamp t10.
A second passes and a User executes UPDATE tbl SET C='data' where key=5. Node B coordinates the request and it is assigned timestamp t6.
User executes the query SELECT C from tbl where key=5. Because the DELETE from Step 1 has a more recent timestamp (t10 > t6), no results are returned.
Note that newer versions of the datastax drivers will start defaulting to use Client Timestamps to have your client application generate and assign timestamps to requests instead of relying on the C* nodes to assign them. datastax java-driver as of 3.0 now defaults to client timestamps (read more about there in 'Client-side generation'). This is very nice if all requests come from the same client, however if you have multiple applications writing to cassandra you now have to worry about keeping your client clocks in sync.
It seems to me that using IF would make the statement possibly fail if re-tried. Therefore, the statement is not idempotent. For instance, given the CQL below, if it fails because of a timeout or system problem and I retry it, then it may not work because another person may have updated the version between retries.
UPDATE users
SET name = 'foo', version = 4
WHERE userid = 1
IF version = 3
Best practices for updates in Cassandra are to make updates idempotent, yet the IF operator is in direct opposition to this. Am I missing something?
If your application is idempotent, then generally you wouldn't need to use the expensive IF clause, since all your clients would be trying to set the same value.
For example, suppose your clients were aggregating some values and writing the result to a roll up table. Each client would calculate the same total and write the same value, so it wouldn't matter if multiple clients wrote to it, or what order they wrote to it, since it would be the same value.
If what you are actually looking for is mutual exclusion, such as keeping a bank balance, then the IF clause could be used. You might read a row to get the current balance, then subtract some money and update the balance only if the balance hadn't changed since you read it. If another client was trying to add a deposit at the same time, then it would fail and would have to try again.
But another way to do that without mutual exclusion is to write each withdrawal and deposit as a separate clustered transaction row, and then calculate the balance as an idempotent result of applying all the transaction rows.
You can use the IF clause for idempotent writes, but it seems pointless. The first client to do the write would succeed and Cassandra would return the value "applied=True". And the next client to try the same write would get back "applied=False, version=4", indicating that the row had already been updated to version 4 so nothing was changed.
This question is more about linerizability(ordering) than idempotency I think. This query uses Paxos to try to determine the state of the system before applying a change. If the state of the system is identical then the query can be retried many times without a change in the results. This provides a weak form of ordering (and is expensive) unlike most Cassandra writes. Generally you should only use CAS operations if you are attempting to record state of a system (rather than a history or log)
Do not use many of these queries if you can help it, the guidelines suggest having only a small percentage of your queries rely on this behavior.
This question is about NoSQL (for instance take cassandra).
Is it true that when you use a NoSQL database without data replication that you have no consistency concerns? Also not in the case of access concurrency?
What happens in case of a partition where the same row has been written in both partitions, possible multiple times? When the partition is gone, which written value is used?
Let's say you use N=5 W=3 R=3. This means you have guaranteed consistency right? How good is it to use this quorum? Having 3 nodes returning the data isn't that a big overhead?
Can you specify on a per query basis in cassandra whether you want the query to have guaranteed consistency? For instance you do an insert query and you want to enforce that all replica's complete the insert before the value is returned by a read operation?
If you have: employees{PK:employeeID, departmentId, employeeName, birthday} and department{PK:departmentID, departmentName} and you want to get the birthday of all employees with a specific department name. Two problems:
you can't ask for all the employees with a given birthday (because you can only query on the primary key)
You can't join the employee and the department column families because joins are impossible.
So what you can do is create a column family:
departmentBirthdays{PK:(departmentName, birthday), [employees-whos-birthday-it-is]}
In that case whenever an employee is fired/hired it has to be removed/added in the departmentBirthdays column family. Is this process something you have to do manually? So you have to manually create queries to update all redundant/denormalized data?
I'll answer this from the perspective of cassandra, coz that's what you seem to be looking at (hardly any two nosql stores are the same!).
For a single node, all operations are in sequence. Concurrency issues can be orthogonal though...your web client may have made a request, and then another, but due to network load, cassandra got the second one first. That may or may not be an issue. There are approaches around such problems, like immutable data. You can also leverage "lightweight transactions".
Cassandra uses last write wins to resolve conflicts. Based on you replication factor and consistency level for your query, this can work well.
Quurom for reads AND writes will give you consistency. There is an edge case..if the coordinator doesn't know a quorum node is down, it sends the write requests, then the write would complete when quorum is re-established. The client in this case would get a timeout and not a failure. The subsequent query may get the stale data, but any query after that will get latest data. This is an extreme edge case, and typically N=5, R=3, W3= will give you full consistency. Reading from three nodes isn't actually that much of an overhead. For a query with R=3, the client would make that request to the node it's connected to (the coordinator). The coordinator will query replicas in parallel (not sequenctially). It willmerge up the results with LWW to get the result (and issue read repairs etc. if needed). As the queries happen in parallel, the overhead is greatly reduced.
Yes.
This is a matter of data modelling. You describe one approach (though partitioning on birthday rather than dept might be better and result in more even distribution of partitions). Do you need the employee and department tables...are they needed for other queries? If not, maybe you just need one. If you denormalize, you'll need to maintain the data manually. In Cassandra 3.0, global indexes will allow you to query on an index without being inefficient (which is the case when using a secondary index without specifying the partition key today). Yes another option is to partition employeed by birthday and do two queries, and do the join in memory in the client. Cassandra queries hitting a partition are very fast, so doing two won't really be that expensive.
Ok so a simple task such as generating a sequential number has caused us an issue in the cloud.
Where you have more than one server it gets harder and harder to guarantee that the allocated number between servers are not clashing.
We are using Azure servers if it helps.
We thought about using the app cache but you cannot guarantee it will be updated between servers.
We are limited to using:
a SQL table with an identity column
or
some peer to peer method between servers
or
use a blob store and utalise the locks to store the nost upto date number. (this could have scaling issues)
I just wondered of anyone has an idea of a solution to resolve this?
Surely its a simple problem and must have been solved by now.
If you can live with a use-case where sometimes the numbers you get from this central location are not always sequential (but guaranteed to be unique) I would suggest considering the following pattern. I've helped an large e-commerce client implement this since they needed unique int PK's to synchronize back to premise:
Create a queue and create a small always-running process that populates this queue with sequential integers (this process should remember which number it generated last and keep replenishing the pool with more numbers once the queue gets close to be empty)
Now, you can have your code first poll the next number from the queue, delete it from the queue and then attempt to save it into the SQL Azure database. In case of failure, all you'll have is a "hole" in your sequential numbers. In scenarios with frequent inserts, you may be saving things out of order to the database (two processes poll from queue, one polls first but saves last, the PK's saved to the database are not sequential anymore)
The biggest downside is that you now have to maintain/monitor a process that replenishes the pool of PK's.
After read this, I would not trust on identity column.
I think the best way is before insert, get the last stored id and increment it by one. (programatically). Another option is create a trigger, but it could be a mass if you'll receive a lot of concurrent requests on DB or if your table have millions of records.
create trigger trigger_name
on table_name
after insert
as
declare #seq int
set #seq = (select max(id) + 1 from table_name)
update table_name
set table_name.id = #seq
from table_name
inner join inserted
on table_name.id = inserted.id
More info:
http://msdn.microsoft.com/en-us/library/windowsazure/ee336242.aspx
If you're worried about scaling the number generation when using blobs, then you can use the SnowMaker library which is available on GitHub and Nuget. It gets around the scale problem by retrieving blocks of ids into a local cache. This guarantees that the Ids are unique, but not necessarily sequential if you have more than one server. I'm not sure if that would achieve what you're after.