One way to use Sandstorm is to run the software on your own server -- we call that self-hosting. This page answers common questions from self-hosters.

How do I log in, if there's a problem with logging in via the web?

If logging into your Sandstorm server over the web isn't working, you can reset your Sandstorm's login providers. Resetting login providers will retain all existing accounts, including account metadata such as who is an admin.

These instructions assume you've installed Sandstorm as root, which is the default recommendation. If not, remove the sudo from the instructions below.

  • Use e.g. ssh to log into the server running Sandstorm.

  • Run this command to generate a token you can use to log in as an admin, for emergency administration.

    sudo sandstorm admin-token
    

This will print a message such as:

  Generated new admin token.

  Please proceed to http://sandstorm.example.com/admin/19bc20df04838fdc03101d898be075cc02de66f2
  in order to access the admin settings page and configure your login system. This token will
  expire in 15 min, and if you take too long, you will have to regenerate a new token with
  `sandstorm admin-token`.
  • Visit the link it printed out, which gives you emergency access to the server's admin panel.

  • From there, configure the login providers of your choosing.

  • Now, log in as yourself. If you log in as the first user that ever signed into this Sandstorm instance, then you will be an admin.

Why does Sandstorm require a wildcard host?

See Why Sandstorm needs a wildcard host.

How do I change the domain name for my Sandstorm server?

If your Sandstorm server is currently online, and you want to change the domain name used to reach it, you'll need to adjust /opt/sandstorm/sandstorm.conf and restart Sandstorm.

Before you change the domain name, read the following notes to avoid common problems.

  • If you're using sandcats free HTTPS, you'll need to disable Sandstorm's built-in HTTPS when you switch to a new domain name. Sandstorm's built-in HTTPS only supports sandcats domains at the moment. We recommend disabling HTTPS and getting Sandstorm working on your domain with just HTTP first, to avoid confusion, and then adding HTTPS in a second step. When you're ready, you can configure nginx or other reverse proxy tools to get HTTPS on your own domain name.

  • Your new Sandstorm domain will need to support wildcard DNS. You should set up your wildcard DNS record before you change Sandstorm's hostname.

  • Changing the server hostname will temporarily disable Google and GitHub login, since those login providers embed an assumption about your server's hostname. You will be able to re-enable the login providers after you switch to the new Sandstorm server hostname; you can use command line access to gain temporary admin access to the server while the login providers are disabled.

To change the domain name, edit /opt/sandstorm/sandstorm.conf. To do that, use ssh or a similar tool to gain access to your server, and run a command such as:

sudo nano /opt/sandstorm/sandstorm.conf

In this file, you should change BASE_URL to your new URL, including http:// or https:// and any port number that you need. You should typically set WILDCARD_HOST to *.{{yourSandstormDomain}}.

The WILDCARD_HOST setting should use the same port number as the BASE_URL. If your service is on port 80 and is a HTTP service, no port number is needed; if your service is on port 443 and is a HTTPS service, no port number is needed. WILDCARD_HOST should not specify a protocol.

In order to disable HTTPS for sandcats, look for a line starting with HTTPS_PORT= and remove it.

Finally, save and exit your text editor (for example with Ctrl-o <ENTER> and Ctrl-x in nano).

Now restart Sandstorm to make the changes take effect. You can typically do that by running:

sudo sandstorm restart

Now visit your server's admin panel to make sure that all settings look OK. Make sure to review the login settings and test that grains launch properly. If you can't get into the admin panel, you can generate an "admin token" on the command line by running this command.

sudo sandstorm admin-token

Why can't I access Sandstorm from the Internet, even though the server is running?

If your sandstorm.conf looks like this:

SERVER_USER=sandstorm
PORT=6080
MONGO_PORT=6081
BIND_IP=127.0.0.1
BASE_URL=http://mydomain.com:6080
WILDCARD_HOST=*.mydomain.com:6080
UPDATE_CHANNEL=dev

then you need to change the BIND_IP value to 0.0.0.0.

(To be pedantic, this the unspecified IPv4 address. For IPv6 compatibility, you may want :: instead. We haven't tested this yet.)

What ports does Sandstorm need open?

If you have a strict firewall around the server running Sandstorm, or you are at home and have to enable "port forwarding" on a home wifi gateway, here is a list of the ports Sandstorm needs. This applies on cloud providers like Amazon EC2, where the defaults allow no inbound traffic.

Default configuration

  • TCP port 6080
  • TCP port 30025

Optionally

  • TCP port 443
  • TCP port 80

What are the minimum hardware requirements?

  • Architecture: amd64 (aka x86_64)
  • RAM: 1 GB
  • Disk space: 5 GB
  • Swap: Enabled, if possible

You can probably get away with less, but we wouldn't advise it. 2GB is vastly better than 1 GB.

Using a virtual machine from Amazon EC2, Google Compute Engine, Linode, Digital Ocean, etc., is fine; just make sure you have a recent Linux kernel. The latest Ubuntu LTS release is an easy and good choice of base operating system.

Sometimes I randomly see a lot of errors across the board, while other times the same functions work fine. What's going on?

Do you have enough RAM? Linux will start randomly killing processes when it's low on RAM. Each grain you have open (or had open in the last couple minutes) will probably consume 50MB-500MB of RAM, depending on the app. We therefore recommend using a server with at least 2GB. If you have less that that, see the next question.

My virtual machine doesn't have that much RAM, what can I do?

It might help to set up swap space. The following commands will set up a file on-disk to use as swap:

dd if=/dev/zero of=/swap.img bs=1M count=1024
mkswap /swap.img
swapon /swap.img

echo /swap.img swap swap defaults 0 0 >> /etc/fstab

Why do you support only Google, GitHub, and passwordless email for login?

Using Google or Github for login results in top-notch security and straightforward federated authentication with very little work. This lets Sandstorm be focused on what it's good at. (We could add Twitter, Facebook, etc. login as well, but we are worried about people forgetting which one they used and ending up with multiple accounts.)

For email logins, we chose to avoid passwords entirely. Passwords have a lot of problems. People choose bad passwords. People -- even smart people -- are often fooled by well-crafted phishing attacks. And, of course, people regularly forget their passwords. In order to deal with these threats, we believe that any password-based login system for Sandstorm must, at the very least, support two-factor authentication and be backed by a human security team who can respond to hijackings. There must also be an automated password reset mechanism which must be well-designed and monitored to avoid attacks. Unfortunately, we don't have these things yet. Moreover, we don't believe that building a secure password login system is the best way for Sandstorm to deliver something interesting to the ecosystem.

Another problem with password login is that it makes federation more complicated. When you federate with your friend's server, how does it authenticate you? Not by password, obviously. Perhaps by OpenID or OAuth, but that is again a thing we would need to implement.

In short, we think these are the most secure options we can provide right now.

A note about when and why we think security is important:

  • For self-hosted Sandstorm servers, we want to provide a secure experience.

  • For public Sandstorm servers supporting a large number of users, account security is essential.

  • For a development instance only accessible to localhost, login security may not be particularly important. You can enable the dev accounts feature to create accounts for testing apps.

Federated login enables tracking, and passwordless email login enables anyone with temporary access to an email account to hijack an account. One way to overcome these problems is by building GPG login so you can create an account based on your public key. You can track progress on that effort in this issue.

Why do I see an error when I try to launch an app, even when the Sandstorm interface works fine?

Sometimes Sandstorm seems to be working fine but can launch no apps. This typically relates to Sandstorm's need for wildcard DNS. If you use HTTPS, you will also need wildcard HTTPS. You can read technical details if you wish.

You might see an error screen like this:

Unable to resolve the server's DNS address, screenshot in Chromium

even when the app management interface seems to work fine:

Skinny Sandstorm admin interface, showing your app instance

You might also have seen a warning from the Sandstorm admin settings area. An error will appear if your configuration fails the self-test. If so, read the Javascript console. Sandstorm will log details in your browser's Javascript console, although those details might be minimal; thankfully, your web browser will typically provide further details about the failed connection in the Javascript console. Here are some hints, based on errors you might see.

  • Cross-Origin Request Blocked: The Same Origin Policy disallows reading the remote resource: This usually means your WILDCARD_HOST and BASE_URL disagree about port number, or the WILDCARD_HOST and BASE_URL disagree about your domain name.

  • Connection refused or net::ERR_CONNECTION_REFUSED: This can occur if your WILDCARD_HOST specifies the wrong domain name or if the WILDCARD_HOST and BASE_URL disagree about port number.

  • HTTPS security error or net::ERR_INSECURE_RESPONSE: This can occur if you are using a self-signed certificate for Sandstorm but have not set up a self-signed CA. Read our docs on self-signed SSL for Sandstorm.

Here are some tips to fix configuration issues that can cause WILDCARD_HOST problems.

  • Make sure the WILDCARD_HOST has valid syntax. In the Sandstorm config file (typically /opt/sandstorm/sandstorm.conf), look for the WILDCARD_HOST config item. Note that this should not have a protocol as part of it but does need a port number if BASE_URL specifies a port number. A valid line might be:
WILDCARD_HOST=*.yourname.sandcats.io:6080
  • Make sure wildcard DNS works for your chosen domain. See also our documentation on wildcard DNS. If setting up wildcard DNS is a hassle for you, consider using our free Sandcats dynamic DNS service for your BASE_URL and WILDCARD_HOST.

  • Make sure wildcard HTTPS works on your server if your server uses HTTPS. Consider using our free Sandcats dynamic DNS and HTTPS service if you are OK choosing a subdomain of sandcats.io as your server's name, e.g. example.sandcats.io. If you are using a self-signed certificate, note that you must import a CA certificate into all web browsers where you want the Sandstorm server to able to be viewed. Web browsers do not show a "OK to continue?" prompt for IFRAMEs, and Sandstorm uses IFRAMEs. Read more in the self-signed SSL guide. You can also read our HTTPS topic guide.

Can I customize the root page of my Sandstorm install?

You can definitely customize the root page of your Sandstorm install. You might have noticed that Oasis has a customized front page.

Customized Oasis front page

This is by contrast with the default, which you can see on our older alpha.sandstorm.io service.

Uncustomized front page

This is achieved by configuring a web page to be displayed in the background behind the login dialog (the home page when logged out). To configure this setting, visit your server's Admin panel screen and click Personalization. You can enter a URL as the Splash URL (experimental).

For security reasons, the page must be hosted within your Sandstorm server's wildcard host (otherwise it will be blocked by Content-Security-Policy). We suggest using a static web publishing app like Hacker CMS to host the content.

This feature is experimental; in particular the style and positioning of the login box is subject to change without notice. Please let us know if you'd like to see it stabilize.

When creating your own page like this, we suggest using the Oasis splash URL as a starting point. Use your browser's DOM inspector to find the IFRAME that is on the background of Oasis. Use its CSS rules to guide your own.

Can Sandstorm use a HTTP proxy for outgoing connections?

Yes. Set the http_proxy and https_proxy environment variables in your systemd service or init script. Right now, this can be used to install apps, but other uses of the proxy are untested. If you discover problems with the HTTP proxy support, please file a bug.

As background, a Sandstorm server uses Internet access to achieve tasks like:

  • Downloading apps to install.

  • Automatically updating itself.

  • Automatically downloading app updates.

  • Updating your IP address on file with the sandcats service.

If your environment requires configuring a HTTP proxy for outbound Internet connectivity, and you are using systemd, then you can edit /etc/systemd/system/sandstorm.service to look like the following. You will then need to run sudo systemctl daemon-reload then sudo systemctl restart sandstorm.

[Unit]
Description=Sandstorm server
After=local-fs.target remote-fs.target network.target
Requires=local-fs.target remote-fs.target network.target

[Service]
Type=forking
ExecStart=/opt/sandstorm/sandstorm start
ExecStop=/opt/sandstorm/sandstorm stop
Environment=http_proxy=http://127.0.0.1:3128/
Environment=https_proxy=http://127.0.0.1:3128/

[Install]
WantedBy=multi-user.target

If you use sysvinit or a different init system, then make whatever similar change results in the http_proxy and https_proxy environment variables being set.

Note that the sandcats.io dynamic DNS protocol requires the ability to send UDP packets to the Internet, so if the system cannot do that, then its IP address will not auto-update. If your IP address does not change frequently, this should be OK.

How do I use Sandstorm with an internal IP address?

Since Sandstorm relies on wildcard DNS, you will need to modify your sandstorm.conf to point at a hostname that resolves to your internal IP address. If your organization cannot provide one, you can either use our free sandcats.io DNS service & HTTPS that uses public IP addresses, or use xip.io's free wildcard DNS for internal IP addresses.

To use xip.io, if your Sandstorm server is at (for example) 10.0.0.2, then you should:

  • Open /opt/sandstorm/sandstorm.conf in your favorite text editor, for example by running sudo nano /opt/sandstorm/sandstorm.conf

  • Find the line containing BASE_URL and modify it to say:

BASE_URL=http://10.0.0.2.xip.io:6080
  • Make sure the port number above corresponds to the port in your PORT=... line.

  • Find the line containing WILDCARD_HOST and modify it to say:

WILDCARD_HOST=*.10.0.0.2.xip.io:6080
  • Make sure the port number is the same as the port number in BASE_URL.

  • Make sure your configuration file does not use the HTTPS_PORT or SANDCATS_BASE_DOMAIN settings, which refer to integrating with the sandcats.io DNS & HTTPS service. If you see them, comment them out or remove them.

#HTTPS_PORT=443
#SANDCATS_BASE_DOMAIN=sandcats.io
  • Save and exit your text editor (for example with Ctrl-o and Ctrl-x in nano).

  • Restart Sandstorm by running this command in a terminal.

sudo sandstorm restart
  • Visit your Sandstorm install at http://10.0.0.2.xip.io/ and make sure it is working OK.

Note that you might not have to do this! For the purpose of this question, an internal IP address is something like 192.168.x.y or 10.x.y.z; see Wikipedia's article on private networks. Many organizations use global IP addresses like 18.x.y.z and rely on their organization firewall to prevent external access; in that case, our free sandcats.io DNS service should work fine.

Keep in mind that xip.io is maintained by the kind and gracious Sam Stephenson, not by a member of the Sandstorm team. If you want to run your own wildcard DNS service similar to xip.io inside your own organization, you can do so by downloading xipd which Sam generously licenses as open source software. You can also set up your own *.sandstorm.example.com subdomain within your organization's domain.

mongod failed to start. What's going on?

If your Sandstorm server isn't working, and you find this text in /opt/sandstorm/var/log/sandstorm.log:

**mongod failed to start. Initial exit code: 100, bailing out now.

then MongoDB is unable to start. Sandstorm operates an embedded MongoDB database instance to store information like what user accounts exist and what permissions they have. Keep the following in mind to address the issue.

  • Your system might not have enough free disk space. Sandstorm requires about 500 MB available space to start successfully.

  • You can read /opt/sandstorm/var/log/mongo.log to find out MongoDB's true error message. Specifically, the file will be in var/log/mongo.log underneath wherever Sandstorm is installed; most Sandstorm installations are at /opt/sandstorm.

  • You might be running into a bug in Sandstorm where it is unable to start MongoDB successfully. If so, this is a bug in Sandstorm that probably affects many, many people, and if you report this issue, we will be grateful; your bug report could lead to a code change that fixes many people's Sandstorm servers. To do that, we need to hear from you.

  • In theory, this error message can occur if your Sandstorm database (stored in /opt/sandstorm/var/mongo) has become corrupted. So far, we have seen no instances of this in the wild. If it does occur, you can likely recover from the situation. Even if there is a problem with the Sandstorm MongoDB instance, note that grain data is safely stored separately, so any grain data would not be affected.

To get further help, please open an issue on GitHub. Please include the most recent 100 lines from the MongoDB log file, if you can.

Installing and running without root privileges

In Sandstorm v0.190 and higher, Sandstorm requires either:

  • The ability to launch Sandstorm as root so it can drop privileges itself, or

  • Unprivileged user namespaces enabled in your system's Linux kernel.

You can ask Sandstorm to do an install as non-root by passing -u to the install script. For example:

curl https://install.sandstorm.io/ > install.sh
bash install.sh -u

If you start Sandstorm as a non-root user, Sandstorm uses unprivileged user namespaces as part of an alternative grain isolation strategy. Some Linux kernels do not have user namespaces available, or make them only available to root, so this article provides advice on how to enable user namespaces. If you use the userns-based sandbox, please be sure to keep up to date with kernel updates.

  • People who don't know how to change a Linux kernel. If you are a customer of a hosting provider, please ask your hosting provider to read this page.

  • Arch Linux users. We suggest starting Sandstorm as root instead to avoid the dependency on user namespaces. In #36969, the Arch Linux kernel maintainer indicated that they are not interested in supporting unprivileged user namespaces. Try the linux-lqx AUR package, or build your own kernel. Read the Arch Linux wiki page on kernels for more information.

  • Docker users. You can launch Sandstorm as root per our recommendations for Docker in our installation guide. Alternatively, if your system needs it, you can use the Debian-specific sysctl, but you will need to provide --cap-add SYS_ADMIN --security-opt seccomp=unconfined to docker run.

  • Debian and Ubuntu users. The Sandstorm install script can use a Debian-specific sysctl to enable unprivileged user namespaces. If you are unable to set it on your system, please run these commands as root and/or upgrade your Linux kernel.

# sysctl -w kernel.unprivileged_userns_clone=1
# echo 'kernel.unprivileged_userns_clone = 1' >> /etc/sysctl.conf"
  • RHEL and CentOS users. We suggest starting Sandstorm as root. If you must use the userns sandbox, note that CentOS/RHEL 7.2 ships a kernel that may be able to support unprivileged usernamespaces. In our testing, further work is needed to properly enable Sandstorm to work within CentOS/RHEL 7.2. If you need help with this, please email support@sandstorm.io.

  • OpenVZ users. If you use an OpenVZ-based hosting provider, please try to run Sandstorm as root. If that does not work, please ask your hosting provider to read our installation documentation and this document.

  • Grsecurity kernel users. Grsecurity seems to block unprivileged user namespaces. Sandstorm should operate properly if you start it as root. If you need use Sandstorm with user namespaces, consider running Sandstorm without Grsecurity. You might be interested to see Sandstorm's track record of successfully blocking exploitation of Linux kernel vulnerabilities.

  • Alpine Linux users. Alpine Linux enables Grsecurity by default. See the previous item and consider using the Alpine vanilla kernel.

As an implementation detail, running install.sh with -d uses user namespaces if available, and enables the Debian-specific sysctl on Debian/Ubuntu systems if possible.

How do I enable WebSockets proxying? or, Why do some apps seem to crash & reload?

Some Sandstorm users find that apps like Telescope and Groove Basin seem to load an initial screen and then refresh the page, in a loop. This is typically a symptom of Sandstorm running behind a reverse proxy that needs WebSockets proxying to be enabled.

For nginx: consult the nginx-example.conf that we provide. Pay special attention to:

  • The map $http_upgrade $connection_upgrade section. You need to add this to the config file for this site.

  • The two proxy_set_header lines relating to Upgrade and Connection.

For apache2: consult the apache-virtualhost.conf that we provide. Pay special attention to the RewriteRule stanzas.

Can I use Let's Encrypt for adding HTTPS to Sandstorm?

Yes, support for Let's Encrypt was added in 0.263. To use Let's Encrypt, your Sandstorm server needs to be able to modify your DNS records. This is made possible by the ACME.js library as long as you use a supported DNS provider (Sandcats.io, Cloudflare, Digital Ocean, DNSimple, Duck DNS, GoDaddy, Gandi, Namecheap, Name.com, AWS Route 53, or Vultr). You can now configure this in the admin panel.

Does Sandstorm run on ARM systems like Raspberry Pi?

Not yet, and there is currently no timeline on adding ARM support to Sandstorm.

We might revisit this in a year or two, once we're happy with the Sandstorm experience on x86-64. You shouldn't plan on ARM support existing at any particular time. If you want Sandstorm on a small computer, we can suggest the following, though we haven't personally tried them.

We are focusing on x86-64 because we only have so much time in the day. If you're a volunteer and interested in tackling the ARM/multi-architecture situation, then please speak up on the related issue.

There are a few obstacles we'd need to overcome for Sandstorm to provide a good experience on ARM.

  • Sandstorm app authors compile their own app packages, so we would need to either ask app authors to cross-compile on their own systems, or we would need to set up a buildd network. If we set up a buildd network, this would come into conflict with the Sandstorm app package signing situation where app authors sign their binaries. They might not have a way to test the ARM binaries that we build on the automated builders, so quality might suffer. This isn't a top priority at the moment within Sandstorm-the-company because it would take time and attention from other things which we think will better serve a greater number of users right now. We would welcome help.

  • Historically, MongoDB doesn't officially support ARM. Thankfully all new versions of MongoDB do support ARM! Read more at this MongoDB issue. Although some Sandstorm apps embed old versions of MongoDB, this is presumably not a big problem anymore.

We hope to be able to provide a more satisfying answer one day. If you don't need ARM in particular, but you want a power-efficient small computer that runs Sandstorm, see the links above.

Is Sandstorm ready for production use? Is it in beta?

Sandstorm is ready for production use, and is used by thousands of users for real work every day. The Sandstorm team itself relies on Sandstorm extensively for our day-to-day work including project management, collaboration, and file management, as do many other individuals and companies. As with any software product, there are plenty of things on our roadmap that we intend to add or improve over time, but the core functionality has been stable for a while.

Can I run Sandstorm on a totally-offline server or airgapped network?

Yes. See: Running Sandstorm offline

Why do I see strange DNS lookups in my server log?

When viewing your server's admin log, you might see DNS-related messages like this. They are harmless and can be ignored. They are probably the result of bots scanning the Internet for security issues that do not affect Sandstorm.

Error: Host "sandstorm-www.notyourdomain.example.com" must have exactly one TXT record.
    at server/pre-meteor.js:728:16
    at QueryReqWrap.asyncCallback [as callback] (dns.js:64:16)
    at QueryReqWrap.onresolve [as oncomplete] (dns.js:216:10)

Error: Error looking up DNS TXT records for host "notyourdomain.example.com": queryTxt ETIMEOUT sandstorm-www.notyourdomain.example.com
    at server/pre-meteor.js:728:16
    at QueryReqWrap.asyncCallback [as callback] (dns.js:64:16)
    at QueryReqWrap.onresolve [as oncomplete] (dns.js:216:10)

The odd log messages are the result of misdirected HTTP requests (probably from bots) and the way Sandstorm's static publishing for external domains works. There's no security issue here, nor any action required by the server admin.

Sandstorm supports publishing static sites at domains that are not part of its "wildcard domain." This allows you to publish a website at a domain like example.com from your Sandstorm server, preserving the domain. To achieve this, you would make a DNS A record for example.com pointing at your Sandstorm server. This ensures that when users navigate to example.com in their browser, the browser sends the request to Sandstorm.

Once Sandstorm receives this request, it needs to know which grain it should dispatch this request to. This is supported by having the user create a DNS TXT record for sandstorm-www.example.com containing the grain's "public ID". If no such DNS record is found, Sandstorm concludes that the domain owner has not indicated a desire to publish a particular Sandstorm grain at the domain from the HTTP request.

What's happening in your log traces is that Sandstorm is receiving HTTP requests for those domains, performing the DNS lookup to see what grain's static publishing folder should be used to answer those requests, and the DNS lookup is either failing (in the case of ETIMEOUT) or returning a response whose format surprises Sandstorm.

Sandstorm logs this event since it could indicate a configuration problem if it occurred for a domain name that should be served from this Sandstorm server.

How do I notify the owner of a resource-hogging process?

  1. Identify the process via top.
  2. cat /proc/<pid>/mountinfo and find the mapping that looks like /opt/sandstorm/var/sandstorm/grains/<grain-id>/sandbox /var. This gives you the grain ID.
  3. sudo sandstorm mongo to get to the Sandstorm mongo shell.
  4. db.grains.find({_id: "<grain-id>"}) to get info about the grain. In particular, the userId is shown.
  5. db.users.find({_id: "<user-id>"}) to get the user record. Look for the list of loginCredentials, and take the first ID shown.
  6. db.users.find({_id: "<login-credential-id>"}) to get info about the user identity.

This should show you the user's profile info including e-mail address, etc.

A very large grain fails to start. How can I fix this issue?

In some cases, grains with a large number of files may fail to start, and the error sandstorm/supervisor.c++:232: overloaded: inotify_add_watch: No space left on device; is logged in the grain log. This is caused by a Linux setting which limits the maximum number of inotify watches a given user account (in this case, sandstorm) may request. Due to how Sandstorm monitors storage usage, this value acts as the limit of the number of files which can exist in simultaneously open grains. Starting with the Linux kernel 5.11, the default value of this setting is dynamically adjusted based on memory, but on older kernels, it may be lower than warranted. The previous default was 8,192.

You can check the setting by executing the following command: cat /proc/sys/fs/inotify/max_user_watches

To increase this limit, add a line like fs.inotify.max_user_watches=32768 to /etc/sysctl.conf and then execute the command sysctl -p to load the configuration.