Time sync is always a good idea if for no other reason than being able to collate logs when debugging. Issues may also result if a client is expecting to reliably compare timestamps, ex. checking if a file has been modified. NTP is very easy to use.
Also of note: kerberos is very sensitive to clock skew if you intend to enable security.
On Jun 5, 2013, at 12:59 AM, Harsh J wrote:
> Skew in time between nodes is highly confusing, even if not a problem
> to the runtime. I'd strongly recommend running a time sync service for
> operational sanity.
> On Wed, Jun 5, 2013 at 8:41 AM, Chris Embree <[EMAIL PROTECTED]> wrote:
>> Yes, NTPD is your best option.
>> On 6/4/13, Ben Kim <[EMAIL PROTECTED]> wrote:
>>> This is very basic & fundamental question.
>>> Is time among all nodes needs to be synced?
>>> I've never even thought of timing in hadoop cluster but recently
>>> experienced my servers going out of sync with time. I know hbase requires
>>> time to by synced due to its timestamp action. But I wonder any of hadoop
>>> functionality requires time sync. Perhaps checkpoint, namenode HA, or
>>> datanode report, etc... hmm
>>> *Benjamin Kim*
>>> *benkimkimben at gmail*
> Harsh J