Since NFS has single point of failure issue. I am thinking to build a storage layer using S3 or Google Cloud Storage as PersistentVolumn in my local k8s cluster.
After a lot of google search, I still cannot find an way. I have tried using s3 fuse to mount volume to local, and then create PV by specifying the hotPath. However, a lot of my pods (for example airflow, jenkins), complained about no write permission, or say "version being changed".
Could someone help figuring out the right way to mount S3 or GCS bucket as a PersistenVolumn from local cluster without using AWS, or GCP.
S3 is not a file system and is not intended to be used in this way.
I do not recommend to use S3 this way, because in my experience any FUSE-drivers very unstable and with I/O operations you will easily ruin you mounted disk and stuck in Transport endpoint is not connected nightmare for you and your infrastructure users. It's also may lead to high CPU usage and RAM leakage.
Useful crosslinks:
How to mount S3 bucket on Kubernetes container/pods?
Amazon S3 with s3fs and fuse, transport endpoint is not connected
How stable is s3fs to mount an Amazon S3 bucket as a local directory
Related
I'd love to expose a Google storage bucket over HDFS to a service.
Service in question is a cluster (SOLR) that can speak only to HDFS, given I have no hadoop (nor need for it), ideally I'd like to have a docker container that would user a Google storage bucket as a backend and expose it's contents via HDFS.
If possible I'd like to avoid mounts (like fuse gcsfs), has anyone done such thing?
I think I could just do mount gcsfs and setup a single node cluster with HDFS, but is there a simpler / more robust way?
Any hints / directions are appreciated.
The Cloud Storage Connector for Hadoop is the tool you might need.
It is not a Docker image but rather an install. Further instructions can be found in the GitHub repository under README.md and INSTALL.md
If it is accessed from AWS S3 you'll need a Service Account with access to Cloud Storage and set the env variable GOOGLE_APPLICATION_CREDENTIALS to /path/to/keyfile.
To use SOLR with GCS, you need indeed a hadoop cluster and you can do that in GCP by creating a dataproc cluster then use the connector mentioned to connect your SOLR solution with GCS. for more info check this SOLR
We would like to process AWS ELB access logs and write them into InfluxDB
to be used for application metrics and monitoring (ex. Grafana).
We configured ELB to store access logs into S3 bucket.
What would be the best way to process those logs and write them to InfluxDB?
What we tried so far was to mount S3 bucket to filesystem using s3fs and then use Telegraf agent for processing. But this approach has some issues: s3fs mounting looks like a hack, and all the files in the bucket are compressed and need to be unzipped before telegraf can process them which makes this task overcomplicated.
Is there any better way?
Thanks,
Oleksandr
Can you just install the telegraf agent on the AWS instance that is generating the logs, and have them sent directly to InfluxDB in real-time?
What is the network bandwidth between Amazon ec2 instances and Amazon S3? I am trying to figure out how long it would take me to copy data from Amazon S3 to Amazon EC2 (and vice versa)
This isn't published information, but ... it's fast.
On smaller instance classes, total Ethernet bandwidth available to the instance can easily be consumed by requests to S3, implying that the limitation isn't the connection to S3.
Provisioning a VPC endpoint for S3 access might also improve throughput to S3.
Bottom line, benchmark it. You will, of course, want to use a bucket that's provisioned in the same region as the instance, for both cost and performance reasons. Data transfer between EC2 and S3 is not billed within a region.
We have a site where users upload files, some of them quite large. We've got multiple EC2 instances and would like to load balance them. Currently, we store the files on an EBS volume for fast access. What's the best way to replicate the files so they can be available on more than one instance?
My thought is that some automatic replication process that uploads the files to S3, and then automatically downloads them to other EC2 instances would be ideal.
EBS snapshots won't work because they replicate the entire volume, and we need to be able to replicate the directories of individual customers on demand.
You could write a shell script that would spawn s3cmd to sync your local filesystem with a S3 bucket whenever a new file is uploaded (or deleted). It would look something like:
s3cmd sync ./ s3://your-bucket/
Depends on what OS you are running on your EC2 instances:
There isn't really any need to add S3 to the mix unless you want to store them there for some other reason (like backup).
If you are running *nix the classic choice might be to run rsync and just sync between instances.
On Windows you could still use rsync or else SyncToy from Microsoft is a simple free option. Otherwise there are probably hundreds of commercial applications in this space...
If you do want to sync to S3 then I would suggest one of the S3 client apps like CloudBerry or JungleDisk, which both have sync functionality...
If you are running Windows it's also worth considering DFS (Distributed File System) which provides replication and is part of Windows Server...
The best way is to use the Amazon Cloud Front service. All of the replication is managed as part of the AWS. Content is served from several different availability zones, but does not require you to have EBS volumes in those zones.
Amazon CloudFront delivers your static and streaming content using a global network of edge locations. Requests for your objects are automatically routed to the nearest edge location, so content is delivered with the best possible performance.
http://aws.amazon.com/cloudfront/
Two ways:
Forget EBS, transfer the files to S3 and use S3 as your file-manager than EBS, add cloudfront and use the common-link everywhere.
Mount S3 bucket on any machines.
1. Amazon CloudFront is a web service for content delivery. It delivers your static and streaming content using a global network of edge locations.
http://aws.amazon.com/cloudfront/
2. You can mount S3 bucket on your linux machine. See below:
s3fs -
http://code.google.com/p/s3fs/wiki/InstallationNotes
- this did work for me. It uses FUSE file-system + rsync to sync the files
in S3. It kepes a copy of all
filenames in the local system & make
it look like a FILE/FOLDER.
That way you can share the S3 bucket on different machines.
I am working on a Java MapReduce app that has to be able to provide an upload service for some pictures from the local machine of the user to an S3 bucket.
The thing is the app must run on an EC2 cluster, so I am not sure how I can refer to the local machine when copying the files. The method copyFromLocalFile(..) needs a path from the local machine which will be the EC2 cluster...
I'm not sure if I stated the problem correctly, can anyone understand what I mean?
Thanks
You might also investigate s3distcp: http://docs.amazonwebservices.com/ElasticMapReduce/latest/DeveloperGuide/UsingEMR_s3distcp.html
Apache DistCp is an open-source tool you can use to copy large amounts of data. DistCp uses MapReduce to copy in a distributed manner—sharing the copy, error handling, recovery, and reporting tasks across several servers. S3DistCp is an extension of DistCp that is optimized to work with Amazon Web Services, particularly Amazon Simple Storage Service (Amazon S3). Using S3DistCp, you can efficiently copy large amounts of data from Amazon S3 into HDFS where it can be processed by your Amazon Elastic MapReduce (Amazon EMR) job flow. You can also use S3DistCp to copy data between Amazon S3 buckets or from HDFS to Amazon S3.
You will need to get the files from the userMachine to at least 1 node before you will be able to use them through a MapReduce.
The FileSystem and FileUtil functions refer to paths either on the HDFS or the local disk of one of the nodes in the cluster.
It cannot reference the user's local system. (Maybe if you did some ssh setup... maybe?)