6. Troubleshooting

6.1. `stork-agent`

This section describes the solutions for some common issues with the Stork agent.

Issue:

A machine is authorized in the Stork server successfully, but there are no daemons.

Description:

The user has installed and started stork-server and stork-agent and authorized the machine. The “Last Refreshed” column has a value on the Machines page and the “Error” column value shows no errors, but the “Daemons” column is still blank. The “Daemons” section on the specific Machine page is also blank.

Solution:

Make sure that the daemons are running:

Kea Control Agent, Kea DHCPv4 server, and/or Kea DHCPv6 server
BIND 9
PowerDNS

Stork looks for the processes named kea-ctrl-agent (for Kea CA), kea-dhcp4 or kea-dhcp6 (for Kea DHCP servers), kea-d2 (for Kea DDNS), named (for BIND 9), or pdns_server (for PowerDNS). Make sure those processes are running and are named appropriately. Use the ps aux (or similar) command to determine whether the processes are running. Currently, Stork does not support detecting off-line services. If BIND 9 or PowerDNS is located in an uncommon location and the Stork agent is unable to detect it, there are several steps that may be helpful. First, enable the DEBUG logging level, so the agent will print more detailed information about locations being checked.

Stork attempts the next four actions to detect named configuration, in order:

Try to parse the -c parameter of the running process;
Use the STORK_AGENT_BIND9_CONFIG environment variable;
Try to parse the output of the named -V command;
Try to find the named.conf file in the default locations.

Stork attempts the next three actions to detect pdns_server configuration, in order: 1. Try to parse the –config-dir parameter of the running process; 2. Use the STORK_AGENT_POWERDNS_CONFIG environment variable; 3. Try to find the pdns.conf file (or a file having a name resulting from the use of the –config-name parameter) in the default locations.

The STORK_AGENT_BIND9_CONFIG or STORK_AGENT_POWERDNS_CONFIG environment variables may be defined to specify the exact locations of the BIND 9 or PowerDNS configuration files.

For BIND 9, make sure that the rndc channel is enabled. By default it is enabled, even if the controls clause is missing. Stork is able to detect default values, so typically there is no administrative action required, unless the rndc channel was explicitly disabled. Make sure the rndc key is readable by the Stork agent.

Also, make sure that BIND 9 has the statistics channel enabled, by adding a statistics-channels entry. Typically, this looks like the following:

statistics-channels {
    inet 127.0.0.1 port 8053 allow { 127.0.0.1; };
};

but it may vary greatly, depending on a given setup. Please consult the BIND 9 Administrator Reference Manual (ARM) for details.

For PowerDNS, make sure that the webserver API is enabled, and that the API key is set. For example:

api=yes
api-key=secret-api-key
webserver=yes
webserver-address=127.0.0.1
webserver-port=8085

Explanation:

If the “Last Refreshed” column has a value, and the “Error” column value has no errors, the communication between stork-server and stork-agent works correctly, which implies that the cause of the problem is between the Stork agent and the daemons. The most likely issue is that none of the Kea/BIND 9/PowerDNS daemons are running. stork-agent communicates with the PowerDNS, BIND 9 and Kea (since Kea version 3.0.0) daemons directly; however, for Kea versions prior to 3.0.0, it communicates with the Kea DHCPv4 and Kea DHCPv6 servers via the Kea Control Agent. If only the “CA” daemon is displayed in the Stork interface, the Kea Control Agent is running, but the DHCP daemons are not.

Issue:

An error occurs after clicking the “Show Zone” button for the selected zone in the zone viewer.

Description:

The “Show Zone” button is clicked for a zone, but the zone viewer shows an error message. The typical error message for PowerDNS server is: “failed to get AXFR credentials for zone 0.0.10.in-addr.arpa: allow-axfr-ips allows neither 127.0.0.1 nor ::1”.

Solution:

Make sure that the DNS server configuration allows zone transfers.

The PowerDNS server has straightforward configuration to enable zone transfers. In fact, the default allow-axfr-ips parameter allows zone transfers from the localhost. If additional IP addresses should be allowed to perform the zone transfer, make sure that the allow-axfr-ips parameter includes 127.0.0.1 or ::1. For example:

allow-axfr-ips=127.0.0.1,::1,172.24.0.10

For BIND 9, the zone transfer is enabled using the allow-transfer parameter. This parameter can be specified at different levels. If there is an issue with a transfer of one of the zones, but not others, it can be an indication that one of the zone-level allow-transfer parameters prohibits the transfer for that zone. Make sure that the allow-transfer permits the transfer. It is also important to note that Stork agent prefers running the zone transfer using a local loopback address (assuming that it finds the DNS server is listening on a local loopback address). In this case, the allow-transfer settings should allow running the zone transfer from the local loopback address. For example:

options {
    listen-on port 53 { 127.0.0.1; 172.24.0.10; };
    allow-transfer { 127.0.0.1; 172.24.0.15; };
};

Note that in the configuration example above, the listen-on parameter explicitly enables the DNS service on the local loopback address. The agent will therefore try to perform the zone transfer using this address. The BIND 9 configuration must also set the allow-transfer parameter to permit zone transfers from the local loopback address, otherwise the aforementioned error will result. Alternatively, the Stork agent supports using the keyword any in the listen-on and allow-transfer parameters; it will attempt zone transfers using the local loopback addresses.

Explanation:

Stork agent uses DNS zone transfer (AXFR) to retrieve the zone data from the monitored DNS servers. If the DNS server configuration does not allow zone transfer, or restricts the zone transfer, such that the Stork agent is not allowed to retrieve the zone data, the zone viewer will report an error.

Issue:: After starting the Stork agent, it gets stuck in an infinite “sleeping” loop.
Description:: stork-agent is running with server support (the --listen-prometheus-only flag is unused). The try to register agent in Stork server message is displayed initially, but the agent only prints the recurring sleeping for 10 seconds before next registration attempt message.
Solution 1:: stork-server is not running. Start the Stork server first and restart the stork-agent daemon.
Solution 2:: The configured server URL in stork-agent is invalid. Correct the URL and restart the agent.

Issue:: stork-agent starts but returns the loaded server cert: /var/lib/stork-agent/certs/cert.pem and key: /var/lib/stork-agent/certs/key.pem message.
Description:: stork-agent runs correctly and its registration is successful. After the started serving Stork Agent message, the agent prints the recurring message about loading server certs. The network traffic analysis to the server reveals that it rejects all packets from the agent (TLS HELLO handshake failed).
Solution:: Re-register the agent to regenerate the certificates, using the stork-agent register command.
Explanation:: The /var/lib/stork-agent/certs/ca.pem file is missing or corrupted. The re-registration removes old files and creates new ones.

Issue:: The cert PEM file is not loaded.
Description:: The agent fails to start and prints an open /var/lib/stork-agent/certs/cert.pem: no such file or directory could not load cert PEM file: /var/lib/stork-agent/certs/cert.pem error message.
Solution:: Re-register the agent to regenerate the certificates, using the stork-agent register command.

Issue:: A connection problem to the DHCP daemon(s) is occurring.
Description:: The agent prints the message problem with connecting to dhcp daemon: unable to forward command to the dhcp6 service: No such file or directory. The server is likely to be offline.
Solution 1:: Try to start the Kea service: systemctl start kea-dhcp4 kea-dhcp6.
Solution 2:: Ensure that the control-socket entry is specified in the Kea DHCP configuration file (kea-dhcp4.conf or kea-dhcp6.conf).
Explanation:: The kea-dhcp4.service or kea-dhcp6.service (depending on the service type in the message) may not be running. If the DHCP daemon is running and operational (it allocates the leases), but the problem is still occurring, inspect the DHCP daemon configuration file (kea-dhcp4.conf or kea-dhcp6.conf). The file must contain the top-level control-socket property with valid content. See the DHCPv4 or DHCPv6 section of the Kea ARM for details. This property is missing by default if Kea is installed from the Debian/Ubuntu repository. To avoid this and similar problems, we recommend using ISC’s official packages on Cloudsmith.

Issue:: stork-agent receives a remote error: tls: certificate required message from Kea.
Description:: The Stork agent and Kea are running, but they cannot establish a connection. The stork-agent log contains the error message mentioned above.
Solution:: Install valid TLS certificates in stork-agent or set the cert-required value in /etc/kea/kea-ctrl-agent.conf to false.
Explanation:: By default, stork-agent does not use TLS when it connects to Kea. If the Kea configuration includes the cert-required value set to true, it requires the Stork agent to use secure connections with valid, trusted TLS certificates. It can be turned off by setting the cert-required value to false when using self-signed certificates, or the Stork agent TLS credentials can be replaced with trusted ones.

Issue:: Kea returns a Kea error response - status: 401, message: Unauthorized message.
Description:: The Stork agent and Kea are running, but they cannot connect. The stork-agent logs contain similar messages: failed to parse responses from Kea: { "result": 401, "text": "Unauthorized" } or Kea error response - status: 401, message: Unauthorized.
Solution:: Check if there are any clients specified in the Kea configuration file in the authentication node.
Explanation:: Kea can be configured to use HTTP Basic Authentication. If it is enabled, the Stork agent will read credentials from the Kea daemon configuration files and use them to authenticate itself. The Stork agent chooses credentials with user name beginning with stork. If there is no such user, the agent will use the first user from the list.

Issue:: During the registration process, stork-agent returns a problem with registering machine: cannot parse address message.
Description:: Stork is configured to use an IPv6 link-local address. The agent prints the try to register agent in Stork server message and then the above error. The agent exists with a fatal status.
Solution:: Use a global IPv6 or an IPv4 address.
Explanation:: IPv6 link-local addresses are not supported by stork-server.

Issue:: A protocol problem occurs during the agent registration.
Description:: During the registration process, stork-agent prints a problem with registering machine: Post "/api/machines": unsupported protocol scheme "" message.
Solution:: The --server-url argument is provided in the wrong format; it must be a canonical URL. It should begin with the protocol (http:// or https://), contain the host (DNS name or IP address; for IPv6 escape them with square brackets), and end with the port (delimited from the host by a colon). For example: http://storkserver:8080.

Issue:: The values in /etc/stork/agent.env were changed, but stork-agent does not notice the changes.
Solution 1:: Restart the daemon.
Solution 2:: Send the SIGHUP signal to the stork-agent process.
Explanation:: stork-agent reads configurations at startup or after receiving the SIGHUP signal.

Issue:: The values in /etc/stork/agent.env were changed and the Stork agent was restarted, but it still uses the default values.
Description:: The agent is running using the stork-agent command. It uses the parameters passed from the command line but ignores the /etc/stork/agent.env file entries. If the agent is running as the systemd daemon, it uses the expected values.
Solution 1:: Load the environment variables from the /etc/stork/agent.env file before running Stork agent. For example, run . /etc/stork/agent.env.
Solution 2:: Run the Stork agent with the --use-env-file switch.
Explanation:: The /etc/stork/agent.env file contains the environment variables, but stork-agent does not automatically load them unless the --use-env-file flag is set; the file must be loaded manually. The default systemd service unit is configured to load this file before starting the agent.

Issue:: Stork shows only the Kea Control Agent on the Daemons page. The running Kea is older than 3.0.0. Stork detects no Kea DHCP servers, although the DHCP daemons are running and allocating leases.
Description:: There is only a single daemon titled “CA” on the Kea Daemons page, but no data about any DHCP daemon or DDNS. The Kea Control Agent and Kea DHCPv4 or Kea DHCPv6 daemon are running and serve leases.
Solution:: The kea-ctrl-agent.conf file is missing the control-sockets property.
Explanation:: Stork detects Kea components using the control socket list from the Kea Control Agent configuration file. The list must be configured properly to allow Stork to send commands to Kea daemons. See the Kea ARM <https://kea.readthedocs.io/en/latest/arm/agent.html#configuration> for details. This property is missing by default if Kea is installed from the Debian/Ubuntu repository. To avoid this and similar problems, we recommend using ISC’s official packages on Cloudsmith.

Issue:: The Stork agent fails to start and returns the following error: failed to load hooks from directory: '[HOOK DIRECTORY]': plugin.Open("[HOOK DIRECTORY]/[FILENAME]"): [HOOK DIRECTORY]/[FILENAME]: file too short or failed to load hooks from directory: '[HOOK DIRECTORY]': plugin.Open("[HOOK DIRECTORY]/[FILENAME]"): [HOOK DIRECTORY]/[FILENAME]: invalid ELF header.
Solution:: Remove the given file from the hook directory.
Explanation:: The file under a given path is not a valid Stork hook.

Issue:: The Stork agent fails to start and returns the following error: Cannot start the Stork Agent: incompatible hook version: 1.0.0.
Solution:: Update the given hook.
Explanation:: The hook is out-of-date and is incompatible with the Stork core application.

Issue:: The Stork agent fails to start and returns the following error: Cannot start the Stork Agent: plugin: symbol Version not found in plugin.
Solution:: Remove or fix the given file.
Explanation:: The hook directory contains the Go plugin, but not the hook; the Go hook does not contain a required symbol.

Issue:: The Stork agent fails to start and returns the following error: Cannot start the Stork Agent: hook library dedicated for another program: Stork Server.
Solution:: Move the incompatible hooks to a separate directory.
Explanation:: The Stork agent requires the hook directory to contain only agent hooks. The error message indicates that the hook directory contains hooks dedicated to the Stork server.

Issue:: The Stork agent starts but the hooks are not loaded. The logs include the following message: Cannot find plugin paths in: /usr/lib/stork-agent/hooks: cannot list hook directory: /usr/lib/stork-agent/hooks: open /usr/lib/stork-agent/hooks: no such file or directory.
Solution:: Create the hook directory or change the path in the configuration.
Explanation:: The hook directory does not exist.

Issue:: The Stork agent fails to start and returns the following error: Cannot start the Stork Agent: open [HOOK DIRECTORY]: permission denied cannot list hook directory.
Solution:: Grant read access to the hook directory to the stork-agent user.
Explanation:: The hook directory is not readable.

Issue:: The Stork agent fails to start and returns the following error: Cannot start the Stork Agent: readdirent [HOOK DIRECTORY]/[FILENAME]: not a directory cannot list hook directory.
Solution:: Change the hook directory path.
Explanation:: A file was found instead of a directory under the given hook directory path.

6.2. `stork-server`

This section describes the solutions for some common issues with the Stork server.

Issue:: The values in /etc/stork/server.env were changed, but stork-server does not notice the changes.
Solution 1:: Restart the daemon.
Solution 2:: Send the SIGHUP signal to the stork-server process.
Explanation:: stork-server reads configurations at startup or after receiving the SIGHUP signal.

Issue:: The values in /etc/stork/server.env were changed and the Stork server was restarted, but it still uses the default values.
Description:: The server is running using the stork-server command. It uses the parameters passed from the command line but ignores the /etc/stork/server.env file entries. If the server is running as the systemd daemon, it uses the expected values.
Solution 1:: Load the environment variables from the /etc/stork/server.env file before running the Stork server. For example, run . /etc/stork/server.env.
Solution 2:: Run the Stork server with the --use-env-file switch.
Explanation:: The /etc/stork/server.env file contains the environment variables, but stork-server does not automatically load them unless the --use-env-file flag is set; the file must be loaded manually. The default systemd service unit is configured to load this file before starting the agent.

Issue:: The server is running but rejects HTTP requests due to a TLS handshake error.
Description:: HTTP requests sent via an Internet browser or tools like curl are rejected. The clients show a message similar to: OpenSSL SSL_write: Broken pipe, errno 32. The Stork server logs contain a TLS handshake error entry with the tls: client didn't provide a certificate description.
Solution 1:: Leave the STORK_REST_TLS_CA_CERTIFICATE environment variable and the --rest-tls-ca flag empty.
Solution 2:: Configure the Internet browser or HTTP tool to use a valid and trusted TLS client certificate. The client certificate must be signed by the authority whose CA certificate was provided in the server configuration.
Explanation:: Providing the STORK_REST_TLS_CA_CERTIFICATE environment variable or the --rest-tls-ca flag turns on TLS client certificate verification. HTTP requests must be assigned with a valid and trusted HTTP certificate, signed by the authority whose CA certificate was provided in the server configuration; otherwise, the request is rejected. This option improves server security by limiting access to only trusted users; it should not be used if there is no CA configured, or if it is desirable to allow login to the Stork server from any computer without prior setup.

Issue:: The Stork server fails to start and returns the following error: permission denied for schema public.
Description:: A fresh installation of the Stork server is made and the database is empty. However, the Stork server does not start and the Stork tool returns an error on the database migration. The logs reveal denied access to the schema public.
Solution 1:: Execute GRANT ALL ON DATABASE stork_db TO stork_user; on the Stork database (replace stork_db and stork_user with the proper names).
Solution 2:: Perform the migration using the Stork tool with the maintenance (e.g., super-admin) database credentials.
Explanation:: In some Postgres installations (by default in Postgres 15 and above), the CREATE permission is only initially granted to the database owner. The Stork server needs this permission to perform the database migration on startup. This permission can be granted manually, or the Stork tool can be used to migrate the schema as the maintenance database user (e.g., super-admin).

Issue:: The Stork server fails to start and returns the following error: Cannot start the Stork Server: failed to load hooks from directory: '[HOOK DIRECTORY]': plugin.Open("[HOOK DIRECTORY]/[FILENAME]"): [HOOK DIRECTORY]/[FILENAME]: file too short or Cannot start the Stork Server: failed to load hooks from directory: '[HOOK DIRECTORY]': plugin.Open("[HOOK DIRECTORY]/[FILENAME]"): [HOOK DIRECTORY]/[FILENAME]: invalid ELF header.
Solution:: Remove the given file from the hook directory.
Explanation:: The file under the given path is not a valid Stork hook.

Issue:: The Stork server fails to start and returns the following error: Cannot start the Stork Server: incompatible hook version: 1.0.0.
Solution:: Update the given hook.
Explanation:: The hook is out-of-date and is incompatible with the Stork core application.

Issue:: The Stork server fails to start and returns the following error: Cannot start the Stork Server: plugin: symbol Version not found in plugin.
Solution:: Remove or fix the given file.
Explanation:: The hook directory contains the Go plugin but not the hook; the Go hook does not contain a required symbol.

Issue:: The Stork server fails to start and returns the following error: Cannot start the Stork Server: hook library dedicated for another program: Stork Agent.
Solution:: Move the incompatible hooks to a separate directory.
Explanation:: The Stork server requires the hook directory to contain only server hooks. The error message indicates that the hook directory contains hooks dedicated to the Stork agent.

Issue:: The Stork server starts but the hooks are not loaded. The logs include the following message: Cannot find plugin paths in: /usr/lib/stork-server/hooks: cannot list hook directory: /usr/lib/stork-server/hooks: open /usr/lib/stork-server/hooks: no such file or directory.
Solution:: Create the hook directory or change the path in the configuration.
Explanation:: The hook directory does not exist.

Issue:: The Stork server fails to start and returns the following error: Cannot start the Stork Server: open [HOOK DIRECTORY]: permission denied cannot list hook directory.
Solution:: Grant read access to the hook directory to the stork-server user.
Explanation:: The hook directory is not readable.

Issue:: The Stork server fails to start and returns the following error: Cannot start the Stork Server: readdirent [HOOK DIRECTORY]/[FILENAME]: not a directory cannot list hook directory.
Solution:: Change the hook directory path.
Explanation:: A file was found instead of a directory under the given hook directory path.

6.3. High Virtual Memory Usage

Stork processes allocate a large amount of virtual memory, which is a common situation for applications written in Golang. The Go runtime uses virtual memory to manage memory efficiently. Virtual memory is not the same as physical memory. The size of the reserved virtual memory depends on the internal implementation details of the Go memory allocator. A high value of virtual memory usage is not alarming, as long as real memory usage is low.

Virtual and physical memory usage can be examined using the ps aux command. Virtual memory usage is displayed in the VSZ column; the RSS column shows physical memory usage.

The usual virtual memory usage of the Stork agent on a machine with 16GB RAM, Go 1.22.4, and Ubuntu 22.04 is about 2.5-3GB. The real memory usage is relatively low, about 10-40MB for Kea deployments with dozens of subnets and host reservations and 40-80MB for deployments with thousands of subnets and host reservations.

References:

6. Troubleshooting

6.1. stork-agent

6.2. stork-server

6.3. High Virtual Memory Usage

6.1. `stork-agent`

6.2. `stork-server`