Prepare for the Leap Second

Posted in: Cassandra, Technical Track

Catch up on how to handle the Leap Second and whether you’re ready for it with our previous updates on the impacts it will have on Cassandra and Linux.

 

Background Information

A leap second will be inserted at the end of June 30, 2015 at 23:59:60 UTC. 

There is a small time difference between the atomic clock and astronomical time (which is based on the rotation of earth). Rotation of earth is slowing down.

To synchronize these times, one second will be added to the atomic clock – a leap second – so that both clocks will synchronize. This will happen on June 30th – July 1st midnight UTC (not in local time, time the same as in GMT time zone). After 23 hours 59 minutes 59 seconds, the time will become 23 hours 59 minutes 60 seconds.

Since this system of correction was implemented in 1972, 26 such leap seconds have been inserted. The most recent one happened on June 30, 2012 at 23:59:60 UTC.

Unlike daylight savings time, which shifts the timezone information and does not alter the underlying UTC time clock on which servers work, a leap-second change is an actual change in the UTC time value. Usually, UTC time is continuous and predictable, but the leap second breaks this normal continuity requiring it to be addressed.

 

What You Need to Know – Summary

The June 2015 leap second event is the addition of one second to the atomic clock on June 30, 2015. Pythian has researched the implications that the upcoming leap second insertion may have and presents the relevant information to its clients and the wider community.

At the operating system level:

  • Windows and AIX servers are not affected by this issue.
  • Linux servers using NTP (network time protocol) may be affected, potentially causing error messages, server hangs or 100% CPU utilization. There are a series of patches and workarounds available, depending upon the needs of the components running on the Linux server.
  • HP-UX servers have NTP patches released in Q2 2015.

For databases and other software components:

  • Java programs are at risk of generating endless error loops, spiking CPU utilization. Patches are available.
  • Databases generally obtain time-stamps from the server OS, so those running on Linux have potential issues. For most, there are no additional corrections necessary.
  • Oracle databases have minimal additional risk. Oracle clustered environments and java-based administration tools should be reviewed and corrective actions taken.
  • Microsoft SQL Server databases have no risk but may expose minor application issues on data granularity and error handling.
  • Open source databases should be reviewed for Java risks. Updated kernels are available.
  • Cisco UCS environments should be reviewed. Patches are available.

Symptoms from the leap second event may persist for up to a day before and after the leap second event, as server NTP service updates are provided.

For all environments, a complete assessment and planning for your systems should be performed. The Pythian team would be pleased to help you perform this assessment and complete the planning necessary to ensure your systems can handle the leap second event in 2015. Get started by reviewing the full Leap Second Report.

 

email

Author

Interested in working with Michael? Schedule a tech call.

No comments

Leave a Reply

Your email address will not be published. Required fields are marked *