I want to run a Tomcat app in Cloudbees. This app accesses some private and confidential properties from the file system. How could I access a file system on Cloudbees? Please note that it should be highly protected, e.g. 700 or similar.
Regards,
Marco
RUN#Cloud platform don't provide a persistent (nor distributed) filesystem. So you can't use it to as canonical store for those files, but need to use an external file store to match your security requirements, and copy them as application is starting (or lazy-load) to java.io.temp directory. As files are stored on RUN#Cloud there is no security issue as your server instance is fully isolated, and files will be deleted after application undeployed/passivated
So you can use Amazon S3 or comparable to store files
Another option is for you to attach properties to the RUN#Cloud instance as configuration parameters, and access them as System properties. See http://wiki.cloudbees.com/bin/view/RUN/Configuration+Parameters
If they data is modest in size - you could consider using properties - using the CLI you can set them using
bees config:set propertyName=value
you can then access that as a System property (for example) in your application. The properties themselves are stored encrypted by cloudbees.
I've actually moved to OpenShift since then and I solved the problem. Thank you for your answers
Related
ASP.NET Core has a great feature to store user settings securely.
It works great on OSX (and Linux, Windows), stores data in JSON:
~/.microsoft/usersecrets/<userSecretsId>/secrets.json
Is it working in the cloud (Cloud Foundry) too? If yes, then were are the values stored?
Secrets exists for safe storage during development by helping prevent sensitive data from being storing in code / checked into source control.
The Secret Manager tool does not encrypt the stored secrets and should not be treated as a trusted store. It is for development purposes only. The keys and values are stored in a JSON configuration file in the user profile directory.
In Cloud you can directly use Environment Variables for that.
I am using database connector component, with vault component to store the database credentials. Now as per the documentation of both components i have created different properties file for each environment to store the encrypted credentials for diff env.
Following is the structure of my mule project
Now the problem with this structure is that i have to build new deployable zip file whenever i have to update the database credentials for any environment.
I need a solution where i can keep all credentials encrypted and centralized and i don't have to create a build every time after updated the credentials, We can afford to restart the server, but building new zip and deploying is really cumbersome.
Second problem we have this approach is a developer needs to know the production db to update it in properties file, this is also a security issue.
Please suggest alternate approach for credentials management for mule projects.
I'm going to recommend you do NOT try to change the secure solution provided to you by MuleSoft. To alleviate the need for packaging and deployment, you would have to extract the properties files outside of the deployment and this would be a huge risk. Regardless of where you store the property files within the deployment if you change the files, you have to package and re-deploy. I see the only solution to your problem as moving the files outside of the deployment and securely storing them. Mule has provided a solution while it may be cumbersome, they are securing these files first with encryption and secondly within the server container. You can move out the property files but you have to provide a custom implementation and you will be assuming great risk to your protected resources.
Set a VM arguement e.g. environment.type=local for local machine on your anypoint studio.
Read this variable in wherever you are reading your properties file in a way that environment type is read dynamically such as below.
" location="classpath:properties/sample-app-${environment.type}.properties" doc:name="Secure Property Placeholder"/>
In order to set the environment type on your production server(or wherever you are using mule runtime), open \conf\wrapper.conf and add the arguement wrapper.java.additional.=-Dserver.type=production. If you already have any property in this file, you may need to set the value of n appropriately. For example 13 or 14.
This way you don't need to generate different deployment artefacts for different environment because correct properties file is picked by using environment specific VM arguement.
We currently have a site running cold fusion 11. In an effort to improve some aspects of security we would like to store all files uploaded by our users on a server separate from our codebase and DB servers.
I'm pretty much starting from scratch here as I wasn't able to find much in my searches so far. What's the best practice for doing this and what cold fusion functions would work for storing and retrieving files from an external source?
I could use some more information to be more helpful. But let's say you have a separate server that stores all your user files on a Windows network. I would use CFContent to serve those files with the file being retrieved over a UNC path.
I'd recommend reading this blog entry of mine on Securely Serving Files via CFContent. Wil, also from CF Webtools, posts one here: Serving File Downloads with ColdFusion
We had a similar issue when we migrated to a Unix platform. Our solution was to mount a file server to the webserver. It's accessed programmatically by ColdFusion as if it's on the same server, but it's inaccessible from the web root (browser). It's worked very smoothly for us.
So, a common practice these days is to put connection strings & passwords as environment variables to avoid their being placed into a file. This is all fine and dandy, but I'm not sure how to make this work when trying to set up a continuous deployment workflow with some configuration management tool such as Salt/Ansible or Chef/Puppet.
Specifically, I have the following questions in environments using the above mentioned configuration management tools:
Where do you store connection strings/passwords/keys separate from codebases?
Do you keep those items in a code-repo of some type (git, etc.)?
Do you use some structure built-in to your tool?
How do you keep those same items secure?
Do you track changes/back-up these items, and if so, how?
In Chef you can
store passwords or API tokens in either encrypted data bags or using chef-vault. They are then decrypted while chef does the provisioning (with encrypted data bags using a shared secret, with chef-vault using the existing PKI of Chef client).
set environment variables when calling external software using the environment parameter of e.g. the execute resource.
not sure, what to write here -- I'd say you don't really manage them. This way you set the variables only for the command that needs it, not e.g. for the whole chef run.
With puppet, the preferred way is probably to store the secrets in Hiera files, which are just plain YAML files. That means that all secrets are stored on the master, separate from the manifest files.
truecrypt virtual encrypted disks are cross-platform and independent of tooling. Mount it read-write to change the secrets in the files it contains, unmount it and then commit/push the encrypted disk image into version control. Mount read-only for automation.
ansible-vault can be used to encrypt sensitive data files. A CI server like Jenkins however is not the safest place to store access credentials. If you add Hashicorp Vault and Ansible Tower/AWX, then you can provide a secure solution for several teams.
In Visual Studio 2012, using publish profiles along with web deploy simplifies the deployments quite a bit. However it still is missing few things or may be I don't know how to use it yet.
I prefer to use the NTLM authentication without storing the username and password (especially) in the publish profiles. How can this be done? If I leave the username and password empty, I am prompted for it. Is there a way like manually modifying the .pubxml files?
Why is the username/password stored in PublishProfileName.pubxml that I have checked in the source control and not in PublishProfileName.pubxml.user that is local to each user? I could at least save the username but obviously don't want that to be checked in.
The Configuration itself is not part of PublishProfileName.pubxml but is stored in PublishProfileName.pubxml.user as LastUsedBuildConfiguration.
Same for the Platform as last point.
I am also missing support for multi-server deployments. I am currently forced to use batch files in addition to Publish Profiles.
EDIT
The command line that works fine for publishing is
MSBuild.Exe MyProject.sln /p:Configuration=QA /p:DeployOnBuild=true;PublishProfile=PublishToQA;AllowUntrustedCertificate=true /p:authType=NTLM /p:UserName=
In this I would like to omit the /p:Configuration=QA if the configuration becomes part of the publish profile itself.
Some answers to your questions.
I prefer to use the NTLM authentication without storing the username and password (especially) in the publish profiles. How can
this be done? If I leave the username and password empty, I am
prompted for it. Is there a way like manually modifying the .pubxml
files?
Your authentication is typically driven by how Web Deploy is hosted. By default if you are using the Web Management Service then you are using IIS users for auth. With IIS users you can control which users have permissions to specific sites/apps. You can configure WMSVC to use windows auth as well though. If you have issues using VS for those scenarios let me know.
If you are using the Remote Agent service to host Web Deploy then in this case you'll be using windows auth.
Why is the username/password stored in PublishProfileName.pubxml that I have checked in the source control and not in
PublishProfileName.pubxml.user that is local to each user? I could
at least save the username but obviously don't want that to be checked
in.
We have another mechanism for you to determine what information is private/shared. With the exception of the password all publish info is shared (and checked in by default). In order to simplify the design you can either have a publish profile which is shared, or one which is not shared at all. There is no in-between in which you have a profile that some fields are shared and other not. Password is special cased here and encrypted on a per-user/per-machine basis in the .pubxml.user file.
If you'd like to have a private publish profile then you can simply not check in the .pubxml file which corresponds to the publish profile. These are stored in the Properties\PublishProfiles (or My Project\PublishProfiles for VB) and just exclude them from the project and don't check the files in. The publish dialog looks for the profiles on disk, not just the ones which are in the project. Everything should continue to work.
We don't support the concept of selectively storing values in the .pubxml.user file. The publish dialog will only store a set number of values in that file. Instead of
The Configuration itself is not part of PublishProfileName.pubxml but is stored in
PublishProfileName.pubxml.user as LastUsedBuildConfiguration.
Same for the Platform as last point.
This was a mistake it should have been stored in the .pubxml file, not the .pubxml.user file. We have since fixed this, but haven't had a chance to release the update yet.
The Configuration property cannot be set in the publish profile. The Configuration property is a core part of the build process. To be more specific, the reason why we didn't call this property Configuration is because the .pubxml file is imported into the definition of the .csproj/.vbproj during a build & publish. Since other properties are defined based on Configuration you cannot change the value once it's been set. I just blogged with way too much detail on this subject at http://sedodream.com/2012/10/27/MSBuildHowToSetTheConfigurationProperty.aspx. This limitation is an MSBuild thing not a publish limitation. For command line you should specify Configuration in the following way:
msbuild.exe myproj.csproj /p:...(other properties)... /p:Configuration=
I am also missing support for multi-server deployments. I am currently forced to use batch files in addition to Publish Profiles.
We don't have direct support for this, but if you expand on your needs I may be able to help. FYI I have an extension which you may be interested in. I have posted a 5 min video to http://sedodream.com/2012/03/14/PackageWebUpdatedAndVideoBelow.aspx.
You are free (and encouraged) to manually edit your pubxml files, so feel free to remove the password.
To switch to NTLM, change AuthType to NTLM in the first PropertyGroup.
Platform and Configuration remain build configuration, the user file just stores them so Visual Studio knows what the last configuration you deployed was.
By multi-server, do you mean a web farm? If so, you might try looking at the Web Farm Framework which basically performs MSDeploy syncs from the primary server to the others.
Alternatively, you could switch to the command line and use postSync to upload and execute a batch file on the remote server that triggers the other deployments from there.