Cassandra's sstableloader too slow in import data - cassandra

I'm using the command line sstableloader to replication datas in the cassandra version 3.11. The process's duration is the 3 hours and 8 minutes to process the amount 356kb. Does this duration is normal independent the amount of data?

sstableload should not take much longer. but you may be throttle your data transfer by default
--throttle 1
(units in MBs). You can speed it up by controlling number of connection per host.
sstableloader --node node_ip --connection-per-host 100
Another reason is if you're using the default and seems very slow. maybe the sstable files loaded compacted while reading them. you need to place your loaded files in directory far from the Cassandra data directory.

Related

How to purge massive data from cassandra when node size is limited

I have a Cassandra cluster (with cassandra v3.11.11) with 3 data centers with replication factor 3. Each node has 800GB nvme but one of the data table is taking up 600GB of data. This results in below from nodetool status:
DataCenter 1:
node 1: 620GB (used) / 800GB (total)
DataCenter 2:
node 1: 610GB (used) / 800GB (total)
DataCenter 3:
node 1: 680GB (used) / 800GB (total)
I cannot install another disk to the server and the server does not have internet connection at all for security reasons. The only thing I can do is to put some small files like scripts into it.
ALL cassandra tables are set up with SizeTieredCompactionStrategy so I end up with very big files ~one single 200GB sstable (full table size 600). Since the cluster is running out of space, I need to free the data which will introduce tombstones. But I have only 120GB of free space left for compaction or garbage collection. That means I can only retain at most 120GB data, and to be safe for the Cassandra process, maybe even less.
I am executed nodetool cleanup so I am not sure if there is more room to free.
Is there any way to free 200GB of data from Cassandra? or I can only retain less than 120 GB data?
(if we remove 200GB data, we have 400GB data left, compaction/gc will need 400+680GB data which is bigger than disk space..)
Thank you!
I personally would start with checking if the whole space is occupied by actual data, and not by snapshots - use nodetool listsnapshots to list them, and nodetool clearsnapshot to remove them. Because if you did snapshot for some reason, then after compaction they are occupying the space as original files were removed.
The next step would be to try to cleanup tombstones & deteled data from the small tables using nodetool garbagecollect, or nodetool compact with -s option to split table into files of different size. For big table I would try to use nodetool compact with --user-defined option on the individual files (assuming that there will be enough space for them). As soon as you free > 200Gb, you can sstablesplit (node should be down!) to split big SSTable into small files (~1-2Gb), so when node starts again, the data will be compacted

Spark hbase bulk load generates more than 15X data

I have Spark dataframe with just 2 columns like { Key| Value}. And this dataframe has 10 million records. I am inserting this into HBase table (has 10 pre-split regions) using bulk load approach from Spark. This works fine and loads the data successfully. When I checked the size table it was like 151GB (453 gb with 3x hadoop replication). I ran major compaction on that table, and table size got reduced to 35GB (105gb with 3x replication).
I am trying to run the same code and same data in a different cluster. But here I have quota limitation of 2TB to my namespace. My process fails while loading HFiles to HBase saying its quota limit exceeded.
I would like to know whether Spark creates much more data files than the required 151GB during the bulk load? If so how to avoid that? or is there better approach to load the same?
The question is that if actual data is around 151gb (before major_compact), then why 2TB size is not enough?

Resolving small file issue in pyspark

I am reading from a partitioned table that has close to 4 billion records.
The files that I am reading from is my source, and I have no control over it to alter the records.
While reading the files through dataframes, for each partition I am creating 2000 files of size less than 2KB. This is because of shuffle partition being set to 2000, to increase the execution speed.
Approach followed to resolve this issue:
I have looped over the HDFS path of the table, post its execution is completed as has created a list with data paths [/dv/hdfs/..../table_name/partition_value=01,/dv/hdfs/..../table_name/partition_value=02..]
For each such path, I have calculated
disk usage and block size from cluster and got the appropriate number of partitions as
no_of_partitions = ceil[disk_usage / block size], and then written the data into another location with the same partition_id such as [/dv/hdfs/..../table2_name/partition_value=01].
Now though this works in reducing the small files to avg block size of 82 MB from 2KB, it is taking about 2.5 mins per partition. With 256 such partitions being available, it is taking more than 10hrs to finish the execution.
Kindly suggest any other method where this could be achieved in less than 2 hrs of time.
Although you have 2000 shuffle partitions you can and should control the output files.
Generating small files in spark is itself a performance degradation for the next read operations.
Now to control small files issue you can do the following:
While writing the dataframe to hdfs repartition it based on the number of partitions and controlling the number of output files per partition
df.repartition(partition_col).write.option("maxRecordsPerFile", 100000).partition_by(partition_col).parquet(path)
This will generate files having 100000 records each in every partition. Hence solving your small files issue. This will improve overall read and write performance of your job.
Hope it helps.

How to retrieve wide partition from cassandra?

We have some large partition in cassandra and I would like to see what caused the large partition. Is there a tool to get the partition data out of cassandra and analyze it ? Right now cqlsh query is timing out even if select single row from the partition.
It can depend on why you get the timeout but theres a some of options.
increase column_index_size_in_kb in your cassandra.yaml to something like 1024 and rebuild the sstables. This works around the object allocation issues a wide partition index introduces.
increasing heap size.
increase read_request_timeout_in_ms
increase key cache size (nodetool setcachecapacity 1000 0 0) then make the read. Watch the read stage until down to zero then try request again. This is hard unless cluster is essentially unused. The read continues even after the timeout, once it finishes reading the index it will cache it so the following read will skip that part (generally the worst) which will speed up significantly
pull raw data from sstable with sstabledump or sstabletools

Cassandra Fast Read Configuration

I have 4 Cassandra nodes with 1 seed in a single data center. I have about 5M records in which Cassandra takes around 4 mins to read where with MySQL, it takes only 17 seconds. So my guess is that there is something wrong in my configuration. So kindly will anyone let me know what configuration attributes so I have to check in Cassandra.yaml.
You may be doing an apples to oranges comparison if you are reading all 5M records from one client.
With MySQL all the data is local and optimized for reads since data is updated in place.
Cassandra is distributed and optimized for writes. Writes are simple appends, but reads are expensive since all the appends need to be read and merged to get the current value of each column.
Since the data is distributed across multiple nodes, there is a lot of overhead of accessing and retrieving the data over the network.
If you were using Spark with Cassandra and loading the data into Spark workers in parallel without shuffling it across the network to a single client, then it would be a more similar comparison.
Cassandra is generally good at ingesting large amounts of data and then working on small slices of it (i.e. partitions) rather than doing table scan operations such as reading the entire table.

Resources