Avamar: Frequently asked questions about the asktime utility
Summary: This article discusses Avamar's asktime utility which is used to perform time synchronization between Avamar nodes.
Instructions
This article discusses the Avamar asktime utility and attempts to answer common questions that are unanswered in the standard documentation.
Q: Do single-node Avamar grids synchronize to external time servers?
A: No, a single-node grid does not have to have a time server at all. If no external time sources are specified, the asktime utility disables the Network Time Protocol (NTP) service Network Time Protocol Daemon (NTPD).
The asktime utility must still be run (or avw_install during an install or upgrade) as more than just NTP must be configured (for example the local time).
Q: Do multinode Avamar grids have external time servers?
A: No, multinode Avamar grids always have internal time servers, which are nodes 0.s and 0.0.
If no external time servers are available, the 0.s or Utility Node's local BIOS clock (hardware clock) becomes the primary time source. (This is not recommended, because PC BIOS clocks are "drifty.")
Q: An external time server was specified, but asktime, avw_install or avw-time never synchronizes to it. What is wrong?
-
The alleged time server does not speak Network Time Protocol (NTP), which uses port 123/
UDP. Several time protocols exist which are not compatible with NTP (such as UNIX daytime). -
An explicit network block, firewall, or a routers access control list (ACL) preventing two-way communication for
UDPprotocol on port 123. -
There is a network connectivity problem somewhere. This is a common issue. Since NTP is one of the first protocols that are configured for external communication, NTP configuration often provides the first indication of a problem with network connectivity.
Q: How to recover during avw_install or asktime when it becomes apparent that the time servers never synchronize?
asktime, but specify no external time servers.
This might proceed past the NTP configuration problem but may proceed later to problems that are related to network connectivity. It is best to understand the root cause of the NTP issue before continuing.
Q: How to defeat avw_install's time configuration altogether on, say, VMware?
A: If running a single-node server under VMware, there is no requirement to specify an external time server.
Q: Why are there so many time synchronization messages in /var/log/messages, a few minutes after the others?
-
Hardware issues: For example, problematic hardware, or interrupts which are hogged by a driver.
-
Network issues: For example, there is so much Avamar backup traffic that
UDPpackets are being dropped. Alternatively, a faulty switch port is causing a high retransmission rate, artificially saturating the network. -
Time server issues: For example, not every time server is synchronized to The One True Time. NTPD likes all its time servers to agree. Otherwise, the NTPD might hop around looking for whatever it considers to be the most reliable source of time.
Time synchronization messages are not harmful, unless the messages are so voluminous that they cause the system logs to consume all space in the /var partition.
Either relieve network congestion, choose better time servers, or fix hardware issues.
Q: Why is NTPD (the NTP service) dying so often?
A: NTP is sensitive to loss of connectivity with time servers. If the network is busy, and UDP packets are being dropped on the floor somewhere in the network, and then NTPD might fail.
There is a cron job on each node that attempts to restart NTPD every 20 minutes.
cron job logs to:
-
/usr/local/avamar/var/cron/ntpd_keepalive_cron.log(Avamar Utility Node) -
/usr/local/avamar/var/ntpd_keepalive_cron.log(Avamar Data Storage Nodes)
Q: How to configure NTP on node types other than utility nodes and data nodes?
A: It is possible to cheat by temporarily putting the IP addresses of those other node types into the probe.xml file. (This should be done before running asktime. Afterwards, remove the extra addresses from probe.xml).