Previously, the Allocate method of the daemon always created a new Lease
object. However, as both the CNI ADD and CHECK commands call Allocate,
and CHECK can be called multiple times, this resulted in multiple Lease
objects being created per pod.
Each of these leases was long lived with its own maintain() loop -
however the daemon only kept track of the most recent one, meaning any
old lease objects remained running forever (and held open their NetNS
files). After a long enough period, this resulted in the system crashing
out with "too many files" or a similar error limits-related error.
This commit updates the behaviour of Allocate() to first check if a
Lease already exists for the given clientID. If none is found, one is
created as before. If a Lease is found, a new Check() mechanism is
called, which simply wakes up the maintain() loop to cause it to check
the status of the lease.
This may fix #329.
Signed-off-by: Emily Shepherd <emily@redcoat.dev>
Currently, hostname is set in the original DHCPREQUEST but not the
renewal. With some DHCP server implementations (such as FreeBSD dhcpd),
this leads to the hostname being cleared in the lease table.
This behavior is inconsistent with other DHCP clients such as dhclient
which set the hostname on the renewal request as well. To fix, use the
same options for acquire and renew.
This is compatible with RFC 2131 (see table 5).
Signed-off-by: Akhil Velagapudi <4@4khil.com>
The client id was constructed differently in the acquire
function compared to the release and renew functions,
which caused the dhcp-server to consider it a different client.
This is now encapsulated in a common function.
Signed-off-by: Fabian Wiesel <fabian.wiesel@sap.com>
First byte of client ID is type, instead of value. See this from
RFC2132:
Code Len Type Client-Identifier
+-----+-----+-----+-----+-----+---
| 61 | n | t1 | i1 | i2 | ...
+-----+-----+-----+-----+-----+---
Signed-off-by: SilverBut <SilverBut@users.noreply.github.com>
Almost every first retry of DHCP will fail due to interface is not up. Add a
fast retry to reduce unnecessary latency.
Signed-off-by: SilverBut <SilverBut@users.noreply.github.com>
The default lease acquisition timeout of 62 seconds is way too long when
running multiple testcases, overrunning the `go test` timeout of 10m. Let
testcases specify a shorter timeout.
Signed-off-by: Dan Williams <dcbw@redhat.com>
Eventually the timeout value will become a CLI argument
The default timeout was nestled all the way in the lease constructor
This commit is the first step in making the timeout configurable by
moving it to the DHCPLease constructor
Signed-off-by: toby lorne <toby@toby.codes>
DHCP REQUEST from DHCP plugin does not include Subnet Mask option parameter (1). Some DHCP servers need that option to be explicit in order to return it in a DHCPACK message.
If not, DHCP plugin returns "DHCP option Subnet Mask not found in DHCPACK" error msg in this type of scenario.
interfaces using dhcp ipam.
Vendor latest dhcp4server, dhcp4client, dhcp4
Added additional tests for new functionality in dhcp2_test.go
Wrap d2g dhcp4client calls with our own which add clientID to packet.
Classless static routes (DHCP option 121) are now processed first.
If CSRs exist, static routes (DHCP option 33) and the gateway default
route are ignored as per RFC 3442.
Updates the spec and plugins to return an array of interfaces and IP details
to the runtime including:
- interface names and MAC addresses configured by the plugin
- whether the interfaces are sandboxed (container/VM) or host (bridge, veth, etc)
- multiple IP addresses configured by IPAM and which interface they
have been assigned to
Returning interface details is useful for runtimes, as well as allowing
more flexible chaining of CNI plugins themselves. For example, some
meta plugins may need to know the host-side interface to be able to
apply firewall or traffic shaping rules to the container.
Add a namespace object interface for somewhat cleaner code when
creating and switching between network namespaces. All created
namespaces are now mounted in /var/run/netns to ensure they
have persistent inodes and paths that can be passed around
between plugin components without relying on the current namespace
being correct.
Also remove the thread-locking arguments from the ns package
per https://github.com/appc/cni/issues/183 by doing all the namespace
changes in a separate goroutine that locks/unlocks itself, instead of
the caller having to track OS thread locking.
This takes some of the machinery from CNI and from the rkt networking
code, and turns it into a library that can be linked into go apps.
Included is an example command-line application that uses the library,
called `cnitool`.
Other headline changes:
* Plugin exec'ing is factored out
The motivation here is to factor out the protocol for invoking
plugins. To that end, a generalisation of the code from api.go and
pkg/plugin/ipam.go goes into pkg/invoke/exec.go.
* Move argument-handling and conf-loading into public API
The fact that the arguments get turned into an environment for the
plugin is incidental to the API; so, provide a way of supplying them
as a struct or saying "just use the same arguments as I got" (the
latter is for IPAM plugins).
Path rewriting causes too many problems when vendoring
vendored code. When CNI code is vendored into rkt,
godep has problems code already vendored by CNI.
The plugin binary actually functions in two modes. The first mode
is a regular CNI plugin. The second mode (when stared with "daemon" arg)
runs a DHCP client daemon. When executed as a CNI plugin, it issues
an RPC request to the daemon for actual processing. The daemon is
required since a DHCP lease needs to be maintained by periodically
renewing it. One instance of the daemon can server arbitrary number
of containers/leases.
