Skip to content

Latest commit

 

History

History
527 lines (363 loc) · 36.4 KB

troubleshooting.md

File metadata and controls

527 lines (363 loc) · 36.4 KB
Raw

Troubleshooting

First of all, check this and ensure that you are deploying to the supported ubuntu version.

Installation Problems

Look here if you have a problem running the installer to set up a new Algo server.

Python version is not supported

The minimum Python version required to run Algo is 3.6. Most modern operation systems should have it by default, but if the OS you are using doesn't meet the requirements, you have to upgrade. See the official documentation for your OS, or manual download it from https://www.python.org/downloads/. Otherwise, you may deploy from docker

Error: "You have not agreed to the Xcode license agreements"

On macOS, you tried to install the dependencies with pip and encountered the following error:

Downloading cffi-1.9.1.tar.gz (407kB): 407kB downloaded
  Running setup.py (path:/private/tmp/pip_build_root/cffi/setup.py) egg_info for package cffi

You have not agreed to the Xcode license agreements, please run 'xcodebuild -license' (for user-level acceptance) or 'sudo xcodebuild -license' (for system-wide acceptance) from within a Terminal window to review and agree to the Xcode license agreements.

    No working compiler found, or bogus compiler options
    passed to the compiler from Python's distutils module.
    See the error messages above.

----------------------------------------
Cleaning up...
Command python setup.py egg_info failed with error code 1 in /private/tmp/pip_build_root/cffi
Storing debug log for failure in /Users/algore/Library/Logs/pip.log

The Xcode compiler is installed but requires you to accept its license agreement prior to using it. Run xcodebuild -license to agree and then retry installing the dependencies.

Error: checking whether the C compiler works... no

On macOS, you tried to install the dependencies with pip and encountered the following error:

Failed building wheel for pycrypto
Running setup.py clean for pycrypto
Failed to build pycrypto
...
copying lib/Crypto/Signature/PKCS1_v1_5.py -> build/lib.macosx-10.6-intel-2.7/Crypto/Signature
running build_ext
running build_configure
checking for gcc... gcc
checking whether the C compiler works... no
configure: error: in '/private/var/folders/3f/q33hl6_x6_nfyjg29fcl9qdr0000gp/T/pip-build-DB5VZp/pycrypto': configure: error: C compiler cannot create executables See config.log for more details
Traceback (most recent call last):
File "", line 1, in
...
cmd_obj.run()
File "/private/var/folders/3f/q33hl6_x6_nfyjg29fcl9qdr0000gp/T/pip-build-DB5VZp/pycrypto/setup.py", line 278, in run
raise RuntimeError("autoconf error")
RuntimeError: autoconf error

You don't have a working compiler installed. You should install the XCode compiler by opening your terminal and running xcode-select --install.

Error: "fatal error: 'openssl/opensslv.h' file not found"

On macOS, you tried to install cryptography and encountered the following error:

build/temp.macosx-10.12-intel-2.7/_openssl.c:434:10: fatal error: 'openssl/opensslv.h' file not found

#include <openssl/opensslv.h>

        ^

1 error generated.

error: command 'cc' failed with exit status 1

----------------------------------------
Cleaning up...
Command /usr/bin/python -c "import setuptools, tokenize;__file__='/private/tmp/pip_build_root/cryptography/setup.py';exec(compile(getattr(tokenize, 'open', open)(__file__).read().replace('\r\n', '\n'), __file__, 'exec'))" install --record /tmp/pip-sREEE5-record/install-record.txt --single-version-externally-managed --compile failed with error code 1 in /private/tmp/pip_build_root/cryptography
Storing debug log for failure in /Users/algore/Library/Logs/pip.log

You are running an old version of pip that cannot download the binary cryptography dependency. Upgrade to a new version of pip by running sudo python3 -m pip install -U pip.

Error: "ansible-playbook: command not found"

You tried to install Algo and you see an error that reads "ansible-playbook: command not found."

You did not finish step 4 in the installation instructions, "Install Algo's remaining dependencies." Algo depends on Ansible, an automation framework, and this error indicates that you do not have Ansible installed. Ansible is installed by pip when you run python3 -m pip install -r requirements.txt. You must complete the installation instructions to run the Algo server deployment process.

Fatal: "Failed to validate the SSL certificate"

You received a message like this:

fatal: [localhost]: FAILED! => {"changed": false, "msg": "Failed to validate the SSL certificate for api.digitalocean.com:443. Make sure your managed systems have a valid CA certificate installed. You can use validate_certs=False if you do not need to confirm the servers identity but this is unsafe and not recommended. Paths checked for this platform: /etc/ssl/certs, /etc/ansible, /usr/local/etc/openssl. The exception msg was: [SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1076).", "status": -1, "url": "https://api.digitalocean.com/v2/regions"}

Your local system does not have a CA certificate that can validate the cloud provider's API. Are you using MacPorts instead of Homebrew? The MacPorts openssl installation does not include a CA certificate, but you can fix this by installing the curl-ca-bundle port with port install curl-ca-bundle. That should do the trick.

Could not fetch URL ... TLSV1_ALERT_PROTOCOL_VERSION

You tried to install Algo and you received an error like this one:

Could not fetch URL https://pypi.python.org/simple/secretstorage/: There was a problem confirming the ssl certificate: [SSL: TLSV1_ALERT_PROTOCOL_VERSION] tlsv1 alert protocol version (_ssl.c:590) - skipping
  Could not find a version that satisfies the requirement SecretStorage<3 (from -r requirements.txt (line 2)) (from versions: )
No matching distribution found for SecretStorage<3 (from -r requirements.txt (line 2))

It's time to upgrade your python.

brew upgrade python3

You can also download python 3.7.x from python.org.

Bad owner or permissions on .ssh

You tried to run Algo and it quickly exits with an error about a bad owner or permissions:

fatal: [104.236.2.94]: UNREACHABLE! => {"changed": false, "msg": "Failed to connect to the host via ssh: Bad owner or permissions on /home/user/.ssh/config\r\n", "unreachable": true}

You need to reset the permissions on your .ssh directory. Run chmod 700 /home/user/.ssh and then chmod 600 /home/user/.ssh/config. You may need to repeat this for other files mentioned in the error message.

The region you want is not available

Algo downloads the regions from the supported cloud providers (other than Microsoft Azure) listed in the first menu using APIs. If the region you want isn't available, the cloud provider has probably taken it offline for some reason. You should investigate further with your cloud provider.

If there's a specific region you want to install to in Microsoft Azure that isn't available, you should file an issue, give us information about what region is missing, and we'll add it.

AWS: SSH permission denied with an ECDSA key

You tried to deploy Algo to AWS and you received an error like this one:

TASK [Copy the algo ssh key to the local ssh directory] ************************
ok: [localhost -> localhost]

PLAY [Configure the server and install required software] **********************

TASK [Check the system] ********************************************************
fatal: [X.X.X.X]: UNREACHABLE! => {"changed": false, "msg": "Failed to connect to the host via ssh: Warning: Permanently added 'X.X.X.X' (ECDSA) to the list of known hosts.\r\nPermission denied (publickey).\r\n", "unreachable": true}

You previously deployed Algo to a hosting provider other than AWS, and Algo created an ECDSA keypair at that time. You are now deploying to AWS which does not support ECDSA keys via their API. As a result, the deploy has failed.

In order to fix this issue, delete the algo.pem and algo.pem.pub keys from your configs directory and run the deploy again. If AWS is selected, Algo will now generate new RSA ssh keys which are compatible with the AWS API.

AWS: "Deploy the template fails" with CREATE_FAILED

You tried to deploy Algo to AWS and you received an error like this one:

TASK [cloud-ec2 : Make a cloudformation template] ******************************
changed: [localhost]

TASK [cloud-ec2 : Deploy the template] *****************************************
fatal: [localhost]: FAILED! => {"changed": true, "events": ["StackEvent AWS::CloudFormation::Stack algopvpn1 ROLLBACK_COMPLETE", "StackEvent AWS::EC2::VPC VPC DELETE_COMPLETE", "StackEvent AWS::EC2::InternetGateway InternetGateway DELETE_COMPLETE", "StackEvent AWS::CloudFormation::Stack algopvpn1 ROLLBACK_IN_PROGRESS", "StackEvent AWS::EC2::VPC VPC CREATE_FAILED", "StackEvent AWS::EC2::VPC VPC CREATE_IN_PROGRESS", "StackEvent AWS::EC2::InternetGateway InternetGateway CREATE_FAILED", "StackEvent AWS::EC2::InternetGateway InternetGateway CREATE_IN_PROGRESS", "StackEvent AWS::CloudFormation::Stack algopvpn1 CREATE_IN_PROGRESS"], "failed": true, "output": "Problem with CREATE. Rollback complete", "stack_outputs": {}, "stack_resources": [{"last_updated_time": null, "logical_resource_id": "InternetGateway", "physical_resource_id": null, "resource_type": "AWS::EC2::InternetGateway", "status": "DELETE_COMPLETE", "status_reason": null}, {"last_updated_time": null, "logical_resource_id": "VPC", "physical_resource_id": null, "resource_type": "AWS::EC2::VPC", "status": "DELETE_COMPLETE", "status_reason": null}]}

Algo builds a Cloudformation template to deploy to AWS. You can find the entire contents of the Cloudformation template in configs/algo.yml. In order to troubleshoot this issue, login to the AWS console, go to the Cloudformation service, find the failed deployment, click the events tab, and find the corresponding "CREATE_FAILED" events. Note that all AWS resources created by Algo are tagged with Environment => Algo for easy identification.

In many cases, failed deployments are the result of service limits being reached, such as "CREATE_FAILED AWS::EC2::VPC VPC The maximum number of VPCs has been reached." In these cases, you must either delete the VPCs from previous deployments, or contact AWS support to increase the limits on your account.

AWS: not authorized to perform: cloudformation:UpdateStack

You tried to deploy Algo to AWS and you received an error like this one:

TASK [cloud-ec2 : Deploy the template] *****************************************
fatal: [localhost]: FAILED! => {"changed": false, "failed": true, "msg": "User: arn:aws:iam::082851645362:user/algo is not authorized to perform: cloudformation:UpdateStack on resource: arn:aws:cloudformation:us-east-1:082851645362:stack/algo/*"}

This error indicates you already have Algo deployed to Cloudformation. Need to delete it first, then re-deploy.

DigitalOcean: error tagging resource

You tried to deploy Algo to DigitalOcean and you received an error like this one:

TASK [cloud-digitalocean : Tag the droplet] ************************************
failed: [localhost] (item=staging) => {"failed": true, "item": "staging", "msg": "error tagging resource '73204383': param is missing or the value is empty: resources"}
failed: [localhost] (item=dbserver) => {"failed": true, "item": "dbserver", "msg": "error tagging resource '73204383': param is missing or the value is empty: resources"}

The error is caused because Digital Ocean changed its API to treat the tag argument as a string instead of a number.

  1. Download doctl
  2. Run doctl auth init; it will ask you for your token which you can get (or generate) on the API tab at DigitalOcean
  3. Once you are authorized on DO, you can run doctl compute tag list to see the list of tags
  4. Run doctl compute tag delete environment:algo --force to delete the environment:algo tag
  5. Finally run doctl compute tag list to make sure that the tag has been deleted
  6. Run algo as directed

Azure: No such file or directory: '/home/username/.azure/azureProfile.json'

TASK [cloud-azure : Create AlgoVPN Server] *****************************************************************************************************************************************************************
An exception occurred during task execution. To see the full traceback, use -vvv. 
The error was: FileNotFoundError: [Errno 2] No such file or directory: '/home/ubuntu/.azure/azureProfile.json'
fatal: [localhost]: FAILED! => {"changed": false, "module_stderr": "Traceback (most recent call last):
File \"/usr/local/lib/python3.6/dist-packages/azure/cli/core/_session.py\", line 39, in load
with codecs_open(self.filename, 'r', encoding=self._encoding) as f:
File \"/usr/lib/python3.6/codecs.py\", line 897, in open\n    file = builtins.open(filename, mode, buffering)
FileNotFoundError: [Errno 2] No such file or directory: '/home/ubuntu/.azure/azureProfile.json'
", "module_stdout": "", "msg": "MODULE FAILURE
See stdout/stderr for the exact error", "rc": 1}

It happens when your machine is not authenticated in the azure cloud, follow this guide to configure your environment

Azure: Deployment Permissions Error

The AAD Application Registration (aka, the 'Service Principal', where you got the ClientId) needs permission to create the resources for the subscription. Otherwise, you will get the following error when you run the Ansible deploy script:

fatal: [localhost]: FAILED! => {"changed": false, "msg": "Resource group create_or_update failed with status code: 403 and message: The client 'xxxxx' with object id 'THE_OBJECT_ID' does not have authorization to perform action 'Microsoft.Resources/subscriptions/resourcegroups/write' over scope '/subscriptions/THE_SUBSCRIPTION_ID/resourcegroups/algo' or the scope is invalid. If access was recently granted, please refresh your credentials."}

The solution for this is to open the Azure CLI and run the following command to grant contributor role to the Service Principal:

az role assignment create --assignee-object-id THE_OBJECT_ID --scope subscriptions/THE_SUBSCRIPTION_ID --role contributor

After this is applied, the Service Principal has permissions to create the resources and you can re-run ansible-playbook main.yml to complete the deployment.

Windows: The value of parameter linuxConfiguration.ssh.publicKeys.keyData is invalid

You tried to deploy Algo from Windows and you received an error like this one:

TASK [cloud-azure : Create an instance].
fatal: [localhost]: FAILED! => {"changed": false,
"msg": "Error creating or updating virtual machine AlgoVPN - Azure Error:
InvalidParameter\n
Message: The value of parameter linuxConfiguration.ssh.publicKeys.keyData is invalid.\n
Target: linuxConfiguration.ssh.publicKeys.keyData"}

This is related to the chmod issue inside /mnt directory which is NTFS. The fix is to place Algo outside of /mnt directory.

Docker: Failed to connect to the host via ssh

You tried to deploy Algo from Docker and you received an error like this one:

Failed to connect to the host via ssh:
Warning: Permanently added 'xxx.xxx.xxx.xxx' (ECDSA) to the list of known hosts.\r\n
Control socket connect(/root/.ansible/cp/6d9d22e981): Connection refused\r\n
Failed to connect to new control master\r\n

You need to add the following to the ansible.cfg in repo root:

[ssh_connection]
control_path_dir=/dev/shm/ansible_control_path

Error: Failed to create symlinks for deploying to localhost

You tried to run Algo and you received an error like this one:

TASK [Create a symlink if deploying to localhost] ********************************************************************
fatal: [localhost]: FAILED! => {"changed": false, "gid": 1000, "group": "ubuntu", "mode": "0775", "msg": "the directory configs/localhost is not empty, refusing to convert it", "owner": "ubuntu", "path": "configs/localhost", "size": 4096, "state": "directory", "uid": 1000}
included: /home/ubuntu/algo-master/playbooks/rescue.yml for localhost

TASK [debug] *********************************************************************************************************
ok: [localhost] => {
    "fail_hint": [
        "Sorry, but something went wrong!",
        "Please check the troubleshooting guide.",
        "https://trailofbits.github.io/algo/troubleshooting.html"
    ]
}

TASK [Fail the installation] *****************************************************************************************

This error is usually encountered when using the local install option and localhost is provided in answer to this question, which is expecting an IP address or domain name of your server:

Enter the public IP address or domain name of your server: (IMPORTANT! This is used to verify the certificate)
[localhost]
:

You should remove the files in /etc/wireguard/ and configs/ as follows:

sudo rm -rf /etc/wireguard/*
rm -rf configs/*

And then immediately re-run ./algo and provide a domain name or IP address in response to the question referenced above.

Wireguard: Unable to find 'configs/...' in expected paths

You tried to run Algo and you received an error like this one:

TASK [wireguard : Generate public keys] ********************************************************************************
[WARNING]: Unable to find 'configs/xxx.xxx.xxx.xxx/wireguard//private/dan' in expected paths.

fatal: [localhost]: FAILED! => {"msg": "An unhandled exception occurred while running the lookup plugin 'file'. Error was a <class 'ansible.errors.AnsibleError'>, original message: could not locate file in lookup: configs/xxx.xxx.xxx.xxx/wireguard//private/dan"}

This error is usually hit when using the local install option on a server that isn't Ubuntu 18.04 or later. You should upgrade your server to Ubuntu 18.04 or later. If this doesn't work, try removing files in /etc/wireguard/ and the configs directories as follows:

sudo rm -rf /etc/wireguard/*
rm -rf configs/*

Then immediately re-run ./algo.

Ubuntu Error: "unable to write 'random state'" when generating CA password

When running Algo, you received an error like this:

TASK [common : Generate password for the CA key] ***********************************************************************************************************************************************************
fatal: [xxx.xxx.xxx.xxx -> localhost]: FAILED! => {"changed": true, "cmd": "openssl rand -hex 16", "delta": "0:00:00.024776", "end": "2018-11-26 13:13:55.879921", "msg": "non-zero return code", "rc": 1, "start": "2018-11-26 13:13:55.855145", "stderr": "unable to write 'random state'", "stderr_lines": ["unable to write 'random state'"], "stdout": "xxxxxxxxxxxxxxxxxxx", "stdout_lines": ["xxxxxxxxxxxxxxxxxxx"]}

This happens when your user does not have ownership of the $HOME/.rnd file, which is a seed for randomization. To fix this issue, give your user ownership of the file with this command:

sudo chown $USER:$USER $HOME/.rnd

Now, run Algo again.

Old Networking Firewall In Place

You may see the following output when attemptint to run ./algo from your localhost:

TASK [Wait until SSH becomes ready...] **********************************************************************************************************************
fatal: [localhost]: FAILED! => {"changed": false, "elapsed": 321, "msg": "Timeout when waiting for search string OpenSSH in xxx.xxx.xxx.xxx:4160"}
included: /home/<username>/algo/algo/playbooks/rescue.yml for localhost

TASK [debug] ************************************************************************************************************************************************
ok: [localhost] => {
    "fail_hint": [
        "Sorry, but something went wrong!",
        "Please check the troubleshooting guide.",
        "https://trailofbits.github.io/algo/troubleshooting.html"
    ]
}

If you see this error then one possible explanation is that you have a previous firewall configured in your cloud hosting provider which needs to be either updated or ideally removed. Removing this can often fix this issue.

Connection Problems

Look here if you deployed an Algo server but now have a problem connecting to it with a client.

I'm blocked or get CAPTCHAs when I access certain websites

This is normal.

When you deploy a Algo to a new cloud server, the address you are given may have been used before. In some cases, a malicious individual may have attacked others with that address and had it added to "IP reputation" feeds or simply a blacklist. In order to regain the trust for that address, you may be asked to enter CAPTCHAs to prove that you are a human, and not a Denial of Service (DoS) bot trying to attack others. This happens most frequently with Google. You can try entering the CAPTCHAs or you can try redeploying your Algo server to a new IP to resolve this issue.

In some cases, a website will block any visitors accessing their site through a cloud hosting provider due to previous, frequent DoS attacks originating from them. In these cases, there is not much you can do except deploy Algo to your own server or another IP that the website has not outright blocked.

I want to change the list of trusted Wifi networks on my Apple device

This setting is enforced on your client device via the Apple profile you put on it. You can edit the profile with new settings, then load it on your device to change the settings. You can use the Apple Configurator to edit and resave the profile. Advanced users can edit the file directly in a text editor. Use the Configuration Profile Reference for information about the file format and other available options. If you're not comfortable editing the profile, you can also simply redeploy a new Algo server with different settings to receive a new auto-generated profile.

Error: "The VPN Service payload could not be installed."

You tried to install the Apple profile on one of your devices and you received an error stating The "VPN Service" payload could not be installed. The VPN service could not be created. Client support for Algo VPN is limited to modern operating systems, e.g. macOS 10.11+, iOS 9+. Please upgrade your operating system and try again.

Little Snitch is broken when connected to the VPN

Little Snitch is not compatible with IPSEC VPNs due to a known bug in macOS and there is no solution. The Little Snitch "filter" does not get incoming packets from IPSEC VPNs and, therefore, cannot evaluate any rules over them. Their developers have filed a bug report with Apple but there has been no response. There is nothing they or Algo can do to resolve this problem on their own. You can read more about this problem in issue #134.

I can't get my router to connect to the Algo server

In order to connect to the Algo VPN server, your router must support IKEv2, ECC certificate-based authentication, and the cipher suite we use. See the ipsec.conf files we generate in the config folder for more information. Note that we do not officially support routers as clients for Algo VPN at this time, though patches and documentation for them are welcome (for example, see open issues for Ubiquiti and pfSense).

I can't get Network Manager to connect to the Algo server

You're trying to connect Ubuntu or Debian to the Algo server through the Network Manager GUI but it's not working. Many versions of Ubuntu and some older versions of Debian bundle a broken version of Network Manager without support for modern standards or the strongSwan server. You must upgrade to Ubuntu 17.04 or Debian 9 Stretch, each of which contain the required minimum version of Network Manager.

Various websites appear to be offline through the VPN

This issue appears occasionally due to issues with MTU size. Different networks may require the MTU to be within a specific range to correctly pass traffic. We made an effort to set the MTU to the most conservative, most compatible size by default but problems may still occur.

If either your Internet service provider or your chosen cloud service provider use an MTU smaller than the normal value of 1500 you can use the reduce_mtu option in the file config.cfg to correspondingly reduce the size of the VPN tunnels created by Algo. Algo will attempt to automatically set reduce_mtu based on the MTU found on the server at the time of deployment, but it cannot detect if the MTU is smaller on the client side of the connection.

If you change reduce_mtu you'll need to deploy a new Algo VPN.

To determine the value for reduce_mtu you should examine the MTU on your Algo VPN server's primary network interface (see below). You might algo want to run tests using ping, both on a local client when not connected to the VPN and also on your Algo VPN server (see below). Then take the smallest MTU you find (local or server side), subtract it from 1500, and use that for reduce_mtu. An exception to this is if you find the smallest MTU is your local MTU at 1492, typical for PPPoE connections, then no MTU reduction should be necessary.

Check the MTU on the Algo VPN server

To check the MTU on your server, SSH in to it, run the command ifconfig, and look for the MTU of the main network interface. For example:

ens4: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>  mtu 1460

The MTU shown here is 1460 instead of 1500. Therefore set reduce_mtu: 40 in config.cfg. Algo should do this automatically.

Determine the MTU using ping

When using ping you increase the payload size with the "Don't Fragment" option set until it fails. The largest payload size that works, plus the ping overhead of 28, is the MTU of the connection.

Example: Test on your Algo VPN server (Ubuntu)
$ ping -4 -s 1432 -c 1 -M do github.com
PING github.com (192.30.253.112) 1432(1460) bytes of data.
1440 bytes from lb-192-30-253-112-iad.github.com (192.30.253.112): icmp_seq=1 ttl=53 time=13.1 ms

--- github.com ping statistics ---
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 13.135/13.135/13.135/0.000 ms

$ ping -4 -s 1433 -c 1 -M do github.com
PING github.com (192.30.253.113) 1433(1461) bytes of data.
ping: local error: Message too long, mtu=1460

--- github.com ping statistics ---
1 packets transmitted, 0 received, +1 errors, 100% packet loss, time 0ms

In this example the largest payload size that works is 1432. The ping overhead is 28 so the MTU is 1432 + 28 = 1460, which is 40 lower than the normal MTU of 1500. Therefore set reduce_mtu: 40 in config.cfg.

Example: Test on a macOS client not connected to your Algo VPN
$ ping -c 1 -D -s 1464 github.com
PING github.com (192.30.253.113): 1464 data bytes
1472 bytes from 192.30.253.113: icmp_seq=0 ttl=50 time=169.606 ms

--- github.com ping statistics ---
1 packets transmitted, 1 packets received, 0.0% packet loss
round-trip min/avg/max/stddev = 169.606/169.606/169.606/0.000 ms

$ ping -c 1 -D -s 1465 github.com
PING github.com (192.30.253.113): 1465 data bytes

--- github.com ping statistics ---
1 packets transmitted, 0 packets received, 100.0% packet loss

In this example the largest payload size that works is 1464. The ping overhead is 28 so the MTU is 1464 + 28 = 1492, which is typical for a PPPoE Internet connection and does not require an MTU adjustment. Therefore use the default of reduce_mtu: 0 in config.cfg.

Change the client MTU without redeploying the Algo VPN

If you don't wish to deploy a new Algo VPN (which is required to incorporate a change to reduce_mtu) you can change the client side MTU of WireGuard clients and Linux IPsec clients without needing to make changes to your Algo VPN.

For WireGuard on Linux, or macOS (when installed with brew), you can specify the MTU yourself in the client configuration file (typically wg0.conf). Refer to the documentation (see man wg-quick).

For WireGuard on iOS and Android you can change the MTU in the app.

For IPsec on Linux you can change the MTU of your network interface to match the required MTU. For example:

sudo ifconfig eth0 mtu 1440

To make the change take affect after a reboot, on Ubuntu 18.04 and later edit the relevant file in the /etc/netplan directory (see man netplan).

Note for WireGuard iOS users

As of WireGuard for iOS 0.0.20190107 the default MTU is 1280, a conservative value intended to allow mobile devices to continue to work as they switch between different networks which might have smaller than normal MTUs. In order to use this default MTU review the configuration in the WireGuard app and remove any value for MTU that might have been added automatically by Algo.

Clients appear stuck in a reconnection loop

If you're using 'Connect on Demand' on iOS and your client device appears stuck in a reconnection loop after switching from WiFi to LTE or vice versa, you may want to try disabling DoS protection in strongSwan.

The configuration value can be found in /etc/strongswan.d/charon.conf. After making the change you must reload or restart ipsec.

Example command:

sed -i -e 's/#*.dos_protection = yes/dos_protection = no/' /etc/strongswan.d/charon.conf && ipsec restart

WireGuard: Clients can connect on Wifi but not LTE

Certain cloud providers (like AWS Lightsail) don't assign an IPv6 address to your server, but certain cellular carriers (e.g. T-Mobile in the United States, EE in the United Kingdom) operate an IPv6-only network. This somehow leads to the Wireguard app not being able to make a connection when transitioning to cell service. Go to the Wireguard app on the device when you're having problems with cell connectivity and select "Export log file" or similar option. If you see a long string of error messages like "Failed to send data packet write udp6 [::]:49727->[2607:7700:0:2a:0:1:354:40ae]:51820: sendto: no route to host then you might be having this problem.

Manually disconnecting and then reconnecting should restore your connection. To solve this, you need to either "force IPv4 connection" if available on your phone, or install an IPv4 APN, which might be available from your carrier tech support. T-mobile's is available for iOS here under "iOS IPv4/IPv6 fix", and here is a walkthrough for Android phones.

IPsec: Difficulty connecting through router

Some routers treat IPsec connections specially because older versions of IPsec did not work properly through NAT. If you're having problems connecting to your AlgoVPN through a specific router using IPsec you might need to change some settings on the router.

Change the "VPN Passthrough" settings

If your router has a setting called something like "VPN Passthrough" or "IPsec Passthrough" try changing the setting to a different value.

Change the default pfSense NAT rules

If your router runs pfSense and a single IPsec client can connect but you have issues when using multiple clients, you'll need to change the Outbound NAT mode to Manual Outbound NAT and disable the rule that specifies Static Port for IKE (UDP port 500). See Outbound NAT in the pfSense Book.

I have a problem not covered here

If you have an issue that you cannot solve with the guidance here, create a new discussion and ask for help. If you think you found a new issue in Algo, file an issue.