ClamAV can be run within a Docker container. This provides isolation from other processes by running it in a containerized environment. If new or unfamiliar with Docker, containers or cgroups see docker.com.
ClamAV image tags on Docker Hub follow these naming conventions.
All images come in two forms:
clamav/clamav:<version>: A release preloaded with signature databases.
Using this container will save the ClamAV project some bandwidth. Use this if you will keep the image around so that you don't download the entire database set every time you start a new container. Updating with FreshClam from existing databases set does not use much data.
clamav/clamav:<version>_base: A release with no signature databases.
Use this container only if you mount a volume in your container under
/var/lib/clamavto persist your signature database databases. This method is the best option because it will reduce data costs for ClamAV and for the Docker registry, but it does require advanced familiarity with Linux and Docker.
Caution: Using this image without mounting an existing database directory will cause FreshClam to download the entire database set each time you start a new container.
There are a selection of tags to help you get the versions you need:
clamav/clamav:<MAJOR.MINOR.PATCH>: This is a tag for a specific image for a given patch version. The "base" version of this image will never change, and the non-base version will only ever be updated to have newer signature databases.
If we need to publish a new image to resolve CVE's in the underlying dependencies, then another image will be created with a build-number suffix.
0.104.2-2_baseis a new image to resolve security issues found in busybox in the
clamav/clamav:<MAJOR.MINOR>: This is a tag for the latest patch version of ClamAV 0.104. When the image for a new patch version is created, this tag will be updated so that it always points to the latest image for ClamAV 0.104.
clamav/clamav:stable: These tags point to the latest stable patch version image. We use the word "stable" to make it clear that these do not track the latest commit in Github. As of 2022-02-15, that makes these equivalent to
0.104_base. When 0.105 is released, these will be updated to track
clamav/clamav:latest: These are the same as
clamav/clamav:stable. They exist because many users expect all images to have a "latest".
clamav/clamav:unstable: These tags point to the latest commit in the
mainbranch on github.com/Cisco-Talos/clamav. Provided something doesn't go wrong, these are updated every evening that something changes in the ClamAV Git repository.
Instead of choosing the specific image for a patch release, choose the tag for a feature release, such as
Only select a "latest" or "stable" tags if you're comfortable with the the risk involved with updating to a new feature release right away without evaluating it first.
_base tag and set up a volume to persist your signature databases. This will save us and you bandwidth. You may choose to set up a container that has the Freshclam daemon enabled, and have multiple others that do not. The ClamD daemon in the all images will occasionally check to see if there are newer signatures in the mounted volume and will reload the databases as needed.
ClamAV uses quite a bit of RAM to load the signature databases into memory. 2GB may be insufficient. Configure your containers to have 4GB of RAM.
The ClamAV Docker images are subject to ClamAV's End-of-Life (EOL) policy. After EOL for a given feature release, those images will no longer be updated and may be unable to download signature updates.
While it is recommended to pull the image from our Docker Hub registry, some may want to build the image locally instead. All that is needed is:
docker build --tag "clamav:TICKET-123" .
in the current directory. This will build the ClamAV image and tag it with the name "clamav:TICKET-123". Any name can generally be used and it is this name that needs to be referred to later when running the image.
clamd in a Docker container, first, an image either has to be built or pulled from a Docker registry.
To pull the ClamAV "unstable" image from Docker Hub, run:
docker pull clamav/clamav:unstable
unstablewith a different version as needed.
To pull and run the official ClamAV images from the Docker Hub registry, try the following command:
docker run \ --interactive \ --tty \ --rm \ --name "clam_container_01" \ clamav/clamav:unstable
The above creates an interactive container with the current TTY connected to it. This is optional but useful when getting started as it allows one to directly see the output and, in the case of
ctrl-c to close the container. The
--rm parameter ensures the container is cleaned up again after it exits and the
--name parameter names the container, so it can be referenced through other (Docker) commands, as several containers of the same image can be started without conflicts.
Note: Pulling is not always required.
docker runwill pull the image if it cannot be found locally.
docker run --pull alwayswill always pull beforehand to ensure the most up-to-date container is being used. Do not use
--pull alwayswith the larger ClamAV images.
Tip: It's common to see
You can run a container using an image built locally (see "Building the ClamAV Image"). Just run:
docker run -it --rm \ --name "clam_container_01" \ clamav:TICKET-123
The virus database in
/var/lib/clamav is by default unique to each container and thus is normally not shared. For simple setups this is fine, where only one instance of
clamd is expected to run in a dockerized environment. However some use cases may want to efficiently share the database or at least persist it across short-lived ClamAV containers.
To do so, you have two options:
Create a Docker volume using the
docker volumecommand. Volumes are completely managed by Docker and are the best choice for creating a persistent database volume.
For example, create a "clam_db" volume:
docker volume create clam_db
Then start one or more containers using this volume. The first container to use a new database volume will download the full database set. Subsequent containers will use the existing databases and may update them as needed:
docker run -it --rm \ --name "clam_container_01" \ --mount source=clam_db,target=/var/lib/clamav \ clamav/clamav:unstable_base
Create a Bind Mount that maps a file system directory to a path within the container. Bind Mounts depend on the directory structure, permissions, and operating system of the Docker host machine.
Run the container with these arguments to mount the a directory from your host environment as a volume in the container.
When doing this, it's best to use the
<version>_baseimage tags so as to save on bandwith. E.g.:
docker run -it --rm \ --name "clam_container_01" \ --mount type=bind,source=/path/to/databases,target=/var/lib/clamav \ clamav/clamav:unstable_base
Disclaimer: When using a Bind Mount, the container's entrypoint script will change ownership of this directory to its "clamav" user. This enables FreshClam and ClamD with the required permissions to read and write to the directory, though these changes will also affect those files on the host.
If you're thinking about running multiple containers that share a single database volume, here are some notes on how this might work.
Scanning files using
clamdscan is possible in various ways with Docker. This section briefly describes them, but the other sections of this document are best read before hand to better understand some of the concepts.
One important aspect is however to realize that Docker by default does not have access to any of the hosts files. And so to scan these within Docker, they need to be mounted with a bind mount to be made accessible.
For example, running the container with these arguments ...
--mount type=bind,source=/path/to/scan,target=/scandir --mount type=bind,source=/path/to/scan,target=/scandir
... would make the hosts file/directory
/path/to/scan available in the container as
/scandir and thus invoking
clamscan would thus be done on
Note that while technically possible to run either scanners via
docker exec this is not described as it is unlikely the container has access to the files to be scanned.
clamscan outside of the Docker container is how normally
clamscan is invoked. To make use of the available shared dockerized resources however, it is possible to expose the virus database and share that for example. E.g. it could be possible to run a Docker container with only the
freshclam daemon running, and share the virus database directory
/var/lib/clamav. This could be useful for file servers for example, where only
clamscan is installed on the host, and
freshclam is managed in a Docker container.
Note: Running the
freshclamdaemon separated from
clamdis less recommended, unless the
clamdsocket is shared with
freshclamwould not be able to inform
clamdof database updates.
clamscan in a Docker container, the Docker container can be invoked as:
docker run -it --rm \ --mount type=bind,source=/path/to/scan,target=/scandir \ clamav/clamav:unstable \ clamscan /scandir
However, this will use whatever signatures are found in the image, which may be slightly out of date. If using
clamscan in this way, it would be best to use a database volume that is up-to-date so that you scan with the latest signatures. E.g.:
docker run -it --rm \ --mount type=bind,source=/path/to/scan,target=/scandir \ --mount type=bind,source=/path/to/databases,target=/var/lib/clamav \ clamav/clamav:unstable_base \ clamscan /scandir
clamdscan can also be run when installed on the host, by connecting to the dockerized
clamd. This can be done by either pointing
clamdscan to the exposed TCP/UDP port or unix socket.
clamdscan is also easily possible, as all that is needed is the shared socket between the two containers. The only cavaet here is to:
- mount the files to be scanned in the container that will run
- mount the files to be scanned in the container that will
clamdscanrun if using
clamdscan --stream. The
--streamoption will be slower, but enables submitting files from a different machine on a network.
docker run -it --rm \ --mount type=bind,source=/path/to/scan,target=/scandir \ --mount type=bind,source=/var/lib/docker/data/clamav/sockets/,target=/run/clamav/ \ clamav/clamav:unstable
docker run -it --rm \ --mount type=bind,source=/path/to/scan,target=/scandir \ --mount type=bind,source=/var/lib/docker/data/clamav/sockets/,target=/run/clamav/ \ clamav/clamav:unstable_base \ clamdscan /scandir
The ClamAV container actually runs both
clamd daemons by default. Optionally available to the container is ClamAV's milter daemon. To control the behavior of the services started within the container, the following flags can be passed to the
docker run command with the
- CLAMAV_NO_CLAMD [true|false] Do not start
clamddaemon is started)
- CLAMAV_NO_FRESHCLAMD [true|false] Do not start the
freshclamdaemon is started)
- CLAMAV_NO_MILTERD [true|false] Do not start the
clamav-milterdaemon is not started)
- CLAMD_STARTUP_TIMEOUT [integer] Seconds to wait for
clamdto start. (default: 1800)
- FRESHCLAM_CHECKS [integer]
freshclamdaily update frequency. (default: once per day)
So to additionally also enable
clamav-milter, the following flag can be added:
Further more, all of the configuration files that live in
/etc/clamav can be overridden by doing a volume-mount to the specific file. The following argument can be added for this purpose. The example uses the entire configuration directory, but this can be supplied multiple times if individual files deem to be replaced.
Note: Even when disabling the
freshclamwill always run at least once during container startup if there is no virus database. While not recommended, the virus database location itself
/var/lib/clamav/could be a persistent Docker volume. This however is slightly more advanced and out of scope of this document.
To connect to a running ClamAV container,
docker exec can be used to run a command on an already running container. To do so, the name needs to be either obtained from
docker ps or supplied during container start via the
--name parameter. The most interesting command in this case can be
docker exec --interactive --tty "clamav_container_01" clamdtop
Alternatively, a shell can be started to inspect and run commands within the container as well.
docker exec --interactive --tty "clamav_container_01" /bin/sh
The default socket for
clamd is located inside the container as
/run/clamav/clamd.sock and can be connected to when exposed via a Docker volume mount. To ensure, that
clamd within the container can freely create and remove the socket, the path for the socket is to be volume-mounted, to expose it for others on the same host to use. The following volume can be used for this purpose. Do ensure that the directory on the host actually exists and clamav inside the container has permission to access it. Caution is required when managing permissions, as incorrect permission could open clamd for anyone on the host system.
With the socket exposed to the host, any other service can now talk to
clamd as well. If for example
clamdtop where installed on the local host, calling
should work just fine. Likewise, running
clamdtop in a different container, but sharing the socket will equally work. While
clamdtop works well as an example here, it is of course important to realize, this can also be used to connect a mail server to
ClamAV in the official Docker images is configured to listen for TCP connections on these ports:
clamav-milter will listen on the above TCP ports, Docker does not expose these by default to the host. Only within containers can these ports be accessed. To expose, or "publish", these ports to the host, and thus potentially over the (inter)network, the
--publish-all) flag to
docker run can be used. While more advanced/secure mappings can be done as per documentation, the basic way is to
--publish [<host_port>:]<container_port> to make the port available to the host.
--publish 13310:3310 \ --publish 7357
The above would thus publish:
13310on the host
7357as a random to the host. The random port can be inspected via
But if you're just running one ClamAV container, you probably will just want to use the default port numbers, which are the same port numbers suggested in the
clamd.conf.sample file provided with ClamAV:
--publish 3310:3310 \ --publish 7357:7357
Warning: Extreme caution is to be taken when using
clamdover TCP as there are no protections on that level. All traffic is un-encrypted. Extra care is to be taken when using TCP communications.
Docker has the ability to run simple
ping checks on services running inside containers. If
clamd is running inside the container, Docker will on occasion send a
clamd on the default port and wait for the pong from
clamd fails to respond, Docker will treat this as an error. The healthcheck results can be viewed with
The performance impact of running
clamd in Docker is negligible. Docker is in essence just a wrapper around Linux's cgroups and cgroups can be thought of as
chroot or FreeBSD's
jail. All code is executed on the host without any translation. Docker does however do some isolation (through cgroups) to isolate the various systems somewhat.
Of course, nothing in life is free, and so there is some overhead. Disk-space being the most prominent one. The Docker container might have some duplication of files for example between the host and the container. Further more, also RAM memory may be duplicated for each instance, as there is no RAM-deduplication. Both of which can be solved on the host however. A filesystem that supports disk-deduplication and a memory manager that does RAM-deduplication.
The base container in itself is already very small ~16 MiB, at the time of thiswriting, this cost is still very tiny, where the advantages are very much worththe cost in general.
The container including the virus database is about ~240 MiB at the time of this writing.
Please, be kind when using 'free' bandwidth, both for the virus databases but also the Docker registry. Try not to download the entire database set or the larger ClamAV database images on a regular basis.
You can run multiple containers that share the same database volume, but be aware that the FreshClam daemons on each would compete to update the databases. Most likely, one would update the databases and trigger its ClamD to load the new databases, while the others would be oblivious to the new databases and would continue with the old signatures until the next ClamD self-check.
This is fine, honestly. It won't take that long before the new signatures are detected by ClamD's self-check and the databases are reloaded automatically.
To reload the databases on all ClamD containers immediately after an update, you could disable the FreshClam daemon when you start the containers. Later, use
docker exec to perform an update and again as needed to have ClamD load updated databases.
Note: This really isn't necessary but you could do this if you wish.
Exactly how you orchestrate this will depend on your environment. You might do something along these lines:
Create a "clam_db" volume, if you don't already have one:
docker volume create clam_db
Start your containers:
docker run -it --rm \ --name "clam_container_01" \ --mount source=clam_db,target=/var/lib/clamav \ --env 'CLAMAV_NO_FRESHCLAMD=true' \ clamav/clamav:0.104_base
Wait for the first one to download the databases (if it's a new database volume). Then start more:
docker run -it --rm \ --name "clam_container_02" \ --mount source=clam_db,target=/var/lib/clamav \ --env 'CLAMAV_NO_FRESHCLAMD=true' \ clamav/clamav:0.104_base
Check for updates, as needed:
docker exec -it clam_container_01 freshclam --on-update-execute=EXIT_1 || \ if [ $? == 1 ]; then \ docker exec -it clam_container_01 clamdscan --reload; \ docker exec -it clam_container_02 clamdscan --reload; \ fi