Implement NTP source info messages to the controller

docs/CONFIG-PROPERTIES.md

@@ -9,6 +9,7 @@
 | timer.metric.diskscan.interval  | integer in seconds | 300 | how frequently device should scan the disk for metrics |
 | timer.location.cloud.interval | integer in seconds | 1 hour | how frequently device reports geographic location information to controller |
 | timer.location.app.interval | integer in seconds | 20 | how frequently device reports geographic location information to applications (to local profile server and to other apps via meta-data server) |
+| timer.ntpsources.interval | integer in seconds | 10 minutes | how frequently device forcibly reports information about NTP sources to which EVE has established a connection for the NTP synchronization. Requests are also sent to the controller if the list of NTP peers or NTP peer fields, such as mode, state, have changed. |
 | timer.send.timeout | timer in seconds | 120 | time for each http/send |
 | timer.dial.timeout | timer in seconds | 10 | maximum time allowed to establish connection |
 | timer.reboot.no.network | integer in seconds | 7 days | reboot after no cloud connectivity |

docs/DEVICE-CONNECTIVITY.md

@@ -278,13 +278,28 @@ Note that proxying of application traffic has to be configured on the applicatio
 
 ## Network Time Protocol (NTP)
 
-EVE always tries to keep its clock synchronized by embedding a NTP Client that synchronizes
+EVE always tries to keep its clock synchronized using NTP daemon (chronyd) that synchronizes
 the system's clock since the boot time.
 This is especially important on the very first boot to ensure that all device certificates
 are generated with valid timestamps. EVE will first try to use NTP servers either received
 from DHCP servers or statically configured in the network override configuration used for
 bootstrapping (see the next section for more info on the network bootstrapping process).
 If no NTP server is provided by DHCP or config, EVE will try to use `pool.ntp.org` by default.
+Every 5 min (see [pkg/pillar/scripts/device-steps.sh](../pkg/pillar/scripts/device-steps.sh)
+for details) EVE fetches NTP server configuration and restarts NTP daemon if
+configuration has been changed.
+
+Also EVE periodically sends information about NTP sources to the controller. NTP sources
+are peers to which EVE has established a connection for the NTP synchronization. The
+reporting period is defied as a runtime configuration property `timer.ntpsources.interval`
+and described in [CONFIG-PROPERTIES.md](CONFIG-PROPERTIES.md). NTP sources updates are also
+sent regardless of the specified interval if the list of NTP peers or NTP peer fields,
+such as mode, state, have changed.
+
+NTP sources reports consist of fields which are well described on the following resources:
+
+- <https://www.ntp.org/documentation/4.2.8-series/ntpq>, "Peer Variables" section
+- <https://chrony-project.org/doc/4.5/chronyc.html>, "Time sources" section
 
 For more info on clock synchronization, refer to [CLOCK-SYNCHRONIZATION.md](CLOCK-SYNCHRONIZATION.md).
 

pkg/pillar/Dockerfile

@@ -127,7 +127,7 @@ ENV BUILD_PKGS patch
 ENV PKGS alpine-baselayout musl-utils libtasn1-progs pciutils yajl xz bash iptables ip6tables iproute2 dhcpcd \
     coreutils dmidecode libbz2 libuuid ipset curl radvd ethtool util-linux e2fsprogs libcrypto1.1 xorriso \
     qemu-img jq e2fsprogs-extra keyutils ca-certificates ip6tables-openrc iptables-openrc ipset-openrc hdparm \
-    libintl libtirpc libblkid zlib rsync
+    libintl libtirpc libblkid zlib rsync chrony
 RUN eve-alpine-deploy.sh
 
 SHELL ["/bin/ash", "-eo", "pipefail", "-c"]

pkg/pillar/cmd/zedagent/handleconfig.go

@@ -83,6 +83,7 @@ type getconfigContext struct {
 	locationCloudTickerHandle interface{}
 	locationAppTickerHandle   interface{}
 	localProfileTickerHandle  interface{}
+	ntpSourcesTickerHandle    interface{}
 	pubDevicePortConfig       pubsub.Publication
 	pubPhysicalIOAdapters     pubsub.Publication
 	devicePortConfig          types.DevicePortConfig

pkg/pillar/cmd/zedagent/handlemetrics.go

@@ -194,8 +194,12 @@ func metricsAndInfoTimerTask(ctx *zedagentContext, handleChannel chan interface{
 
 			locConfig := ctx.getconfigCtx.sideController.locConfig
 			if locConfig != nil {
-				// Publish all info by timer only for LOC
-				triggerPublishAllInfo(ctx, LOCDest)
+				// Publish all info by timer only for LOC. LOC is
+				// always special due its volatile nature, so set
+				// @ForceSend to be sure request will be send
+				// regardless of any checks applied to a request
+				// (check handlentp.go for details).
+				triggerPublishAllInfo(ctx, LOCDest|ForceSend)
 			}
 
 		case <-stillRunning.C: