Articles in this section

Why are a few records shipped to a remote DC after the XDR force linkdown info command has been used?

Detail

When the force-linkdown xdr-command  info command has been used to simulate the loss of a network link to a remote data center, tracking the XDR related metrics indicates that a few records are shipped to the remote destination after the command is issued. Why does this happen?

Answer

This is expected behaviour because the force-linkdown option, when used with the  xdr-command  info command, mimics the behaviour of XDR in a situation where a link to a remote DC is really down.

The sequence is as follows:

  • The link to the remote DC is marked as suspicious.
  • XDR shipping continues.
  • There is a 30 second grace period (to give the link time to recover in the event of a temporary problem.)
  • After 30 seconds the link is checked again.
  • If the link is really down or it has been forced down with the xdr-command info command shipping stops.

It is therefore evident that XDR will continue to ship records during the 30 second grace period after the force linkdown has been issued.


Notes

  • The xdr-command  info command is used to force the link between datacenters down when there are multiple XDR destinations and only 1 requires maintenance.
  • When there is a single XDR destination the enable-xdr  configuration parameter can be changed dynamically to pause XDR shipping while still tracking changes in the digest log.
  • In Aerospike server version 5.0 and above with a re-written XDR, such command will not be necessary as shipping to different DCs will be done independently and rewinding to force reshipping at a point in time will be possible.

Applies To Earliest Version

Pre 4.9

Applies To Latest Version

4.9
Was this article helpful?
0 out of 0 found this helpful