You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@geode.apache.org by Vahram Aharonyan <va...@vmware.com> on 2018/07/06 12:30:50 UTC
Time difference between cluster members
Hi All,
Could someone help to understand whether there is some threshold on time difference between Geode cluster members if they have the same time zone? Or all members should be in sync from time point of view for normal functioning?
Thanks,
Vahram.
Re: Time difference between cluster members
Posted by Anilkumar Gingade <ag...@pivotal.io>.
Vahram,
You are right about the doc. The timestamp is used when resolving the
conflicts between WAN events; within the cluster it does not use timestamp.
-Anil.
On Thu, Jul 19, 2018 at 6:31 AM Vahram Aharonyan <va...@vmware.com>
wrote:
> Hi Anil,
>
>
>
> I was referring to timestamp due to the first statement of this
> documentation-
> https://geode.apache.org/docs/guide/12/developing/distributed_regions/how_region_versioning_works.html#topic_C5B74CCDD909403C815639339AA03758
> :
>
>
>
> “Each region stores version and timestamp information for use in conflict
> detection. Geode members use the recorded information to detect and resolve
> conflicts consistently before applying a distributed update.”
>
>
>
> Although going now after re-reading that I do see that apart from that
> statement timestamp is not mentioned at all.
>
>
>
> Thanks,
>
> Vahram.
>
>
>
> *From:* Anilkumar Gingade <ag...@pivotal.io>
> *Sent:* Wednesday, July 18, 2018 9:04 PM
> *To:* user@geode.apache.org
> *Subject:* Re: Time difference between cluster members
>
>
>
> >> could this time difference create some inconsistence in case of
> replicated regions?
>
> No it doesn't. The consistency check mechanism doesn't use timestamp
> within the cluster.
>
>
>
> -Anil.
>
>
>
>
>
> On Wed, Jul 18, 2018 at 6:35 AM Vahram Aharonyan <va...@vmware.com>
> wrote:
>
> Hi Bruce, Anthony,
>
>
>
> Thank you very much for your replies.
>
> One more question – could this time difference create some inconsistence
> in case of replicated regions? Will this affect conflict checking that
> happens before region updates to detect and resolve concurrent and
> out-of-order updates?
>
>
>
> Thanks,
>
> Vahram.
>
>
>
> *From:* Bruce Schuchardt <bs...@pivotal.io>
> *Sent:* Wednesday, July 11, 2018 7:52 PM
> *To:* user@geode.apache.org
> *Subject:* Re: Time difference between cluster members
>
>
>
> The only time-sync thing I can think of is tombstone expiration. The
> concurrency control mechanism creates tombstones when entries in the cache
> are destroyed. While there is a distributed garbage collection mechanism
> to take care of removing tombstones there is also a local check to see if a
> tombstone has expired, so you don't want to have big clock skews between
> your servers.
>
>
>
> On 7/11/18 2:34 AM, Vahram Aharonyan wrote:
>
> Hi All,
>
>
>
> Any input on this would be highly appreciated.
>
>
>
> Thanks,
>
> Vahram.
>
>
>
> *From:* Vahram Aharonyan <va...@vmware.com> <va...@vmware.com>
> *Sent:* Friday, July 6, 2018 4:31 PM
> *To:* user@geode.apache.org
> *Subject:* Time difference between cluster members
>
>
>
> Hi All,
>
>
>
> Could someone help to understand whether there is some threshold on time
> difference between Geode cluster members if they have the same time zone?
> Or all members should be in sync from time point of view for normal
> functioning?
>
>
>
> Thanks,
>
> Vahram.
>
>
>
>
RE: Time difference between cluster members
Posted by Vahram Aharonyan <va...@vmware.com>.
Hi Anil,
I was referring to timestamp due to the first statement of this documentation- https://geode.apache.org/docs/guide/12/developing/distributed_regions/how_region_versioning_works.html#topic_C5B74CCDD909403C815639339AA03758:
“Each region stores version and timestamp information for use in conflict detection. Geode members use the recorded information to detect and resolve conflicts consistently before applying a distributed update.”
Although going now after re-reading that I do see that apart from that statement timestamp is not mentioned at all.
Thanks,
Vahram.
From: Anilkumar Gingade <ag...@pivotal.io>
Sent: Wednesday, July 18, 2018 9:04 PM
To: user@geode.apache.org
Subject: Re: Time difference between cluster members
>> could this time difference create some inconsistence in case of replicated regions?
No it doesn't. The consistency check mechanism doesn't use timestamp within the cluster.
-Anil.
On Wed, Jul 18, 2018 at 6:35 AM Vahram Aharonyan <va...@vmware.com>> wrote:
Hi Bruce, Anthony,
Thank you very much for your replies.
One more question – could this time difference create some inconsistence in case of replicated regions? Will this affect conflict checking that happens before region updates to detect and resolve concurrent and out-of-order updates?
Thanks,
Vahram.
From: Bruce Schuchardt <bs...@pivotal.io>>
Sent: Wednesday, July 11, 2018 7:52 PM
To: user@geode.apache.org<ma...@geode.apache.org>
Subject: Re: Time difference between cluster members
The only time-sync thing I can think of is tombstone expiration. The concurrency control mechanism creates tombstones when entries in the cache are destroyed. While there is a distributed garbage collection mechanism to take care of removing tombstones there is also a local check to see if a tombstone has expired, so you don't want to have big clock skews between your servers.
On 7/11/18 2:34 AM, Vahram Aharonyan wrote:
Hi All,
Any input on this would be highly appreciated.
Thanks,
Vahram.
From: Vahram Aharonyan <va...@vmware.com>
Sent: Friday, July 6, 2018 4:31 PM
To: user@geode.apache.org<ma...@geode.apache.org>
Subject: Time difference between cluster members
Hi All,
Could someone help to understand whether there is some threshold on time difference between Geode cluster members if they have the same time zone? Or all members should be in sync from time point of view for normal functioning?
Thanks,
Vahram.
Re: Time difference between cluster members
Posted by Anilkumar Gingade <ag...@pivotal.io>.
>> could this time difference create some inconsistence in case of
replicated regions?
No it doesn't. The consistency check mechanism doesn't use timestamp within
the cluster.
-Anil.
On Wed, Jul 18, 2018 at 6:35 AM Vahram Aharonyan <va...@vmware.com>
wrote:
> Hi Bruce, Anthony,
>
>
>
> Thank you very much for your replies.
>
> One more question – could this time difference create some inconsistence
> in case of replicated regions? Will this affect conflict checking that
> happens before region updates to detect and resolve concurrent and
> out-of-order updates?
>
>
>
> Thanks,
>
> Vahram.
>
>
>
> *From:* Bruce Schuchardt <bs...@pivotal.io>
> *Sent:* Wednesday, July 11, 2018 7:52 PM
> *To:* user@geode.apache.org
> *Subject:* Re: Time difference between cluster members
>
>
>
> The only time-sync thing I can think of is tombstone expiration. The
> concurrency control mechanism creates tombstones when entries in the cache
> are destroyed. While there is a distributed garbage collection mechanism
> to take care of removing tombstones there is also a local check to see if a
> tombstone has expired, so you don't want to have big clock skews between
> your servers.
>
>
>
> On 7/11/18 2:34 AM, Vahram Aharonyan wrote:
>
> Hi All,
>
>
>
> Any input on this would be highly appreciated.
>
>
>
> Thanks,
>
> Vahram.
>
>
>
> *From:* Vahram Aharonyan <va...@vmware.com> <va...@vmware.com>
> *Sent:* Friday, July 6, 2018 4:31 PM
> *To:* user@geode.apache.org
> *Subject:* Time difference between cluster members
>
>
>
> Hi All,
>
>
>
> Could someone help to understand whether there is some threshold on time
> difference between Geode cluster members if they have the same time zone?
> Or all members should be in sync from time point of view for normal
> functioning?
>
>
>
> Thanks,
>
> Vahram.
>
>
>
RE: Time difference between cluster members
Posted by Vahram Aharonyan <va...@vmware.com>.
Hi Bruce, Anthony,
Thank you very much for your replies.
One more question - could this time difference create some inconsistence in case of replicated regions? Will this affect conflict checking that happens before region updates to detect and resolve concurrent and out-of-order updates?
Thanks,
Vahram.
From: Bruce Schuchardt <bs...@pivotal.io>
Sent: Wednesday, July 11, 2018 7:52 PM
To: user@geode.apache.org
Subject: Re: Time difference between cluster members
The only time-sync thing I can think of is tombstone expiration. The concurrency control mechanism creates tombstones when entries in the cache are destroyed. While there is a distributed garbage collection mechanism to take care of removing tombstones there is also a local check to see if a tombstone has expired, so you don't want to have big clock skews between your servers.
On 7/11/18 2:34 AM, Vahram Aharonyan wrote:
Hi All,
Any input on this would be highly appreciated.
Thanks,
Vahram.
From: Vahram Aharonyan <va...@vmware.com>
Sent: Friday, July 6, 2018 4:31 PM
To: user@geode.apache.org<ma...@geode.apache.org>
Subject: Time difference between cluster members
Hi All,
Could someone help to understand whether there is some threshold on time difference between Geode cluster members if they have the same time zone? Or all members should be in sync from time point of view for normal functioning?
Thanks,
Vahram.
Re: Time difference between cluster members
Posted by Bruce Schuchardt <bs...@pivotal.io>.
The only time-sync thing I can think of is tombstone expiration. The
concurrency control mechanism creates tombstones when entries in the
cache are destroyed. While there is a distributed garbage collection
mechanism to take care of removing tombstones there is also a local
check to see if a tombstone has expired, so you don't want to have big
clock skews between your servers.
On 7/11/18 2:34 AM, Vahram Aharonyan wrote:
>
> Hi All,
>
> Any input on this would be highly appreciated.
>
> Thanks,
>
> Vahram.
>
> *From:* Vahram Aharonyan <va...@vmware.com>
> *Sent:* Friday, July 6, 2018 4:31 PM
> *To:* user@geode.apache.org
> *Subject:* Time difference between cluster members
>
> Hi All,
>
> Could someone help to understand whether there is some threshold on
> time difference between Geode cluster members if they have the same
> time zone? Or all members should be in sync from time point of view
> for normal functioning?
>
> Thanks,
>
> Vahram.
>
Re: Time difference between cluster members
Posted by Anthony Baker <ab...@pivotal.io>.
Best practice is definitely to run a time sync service like NTP on all cluster members.
> On Jul 11, 2018, at 2:34 AM, Vahram Aharonyan <va...@vmware.com> wrote:
>
> Hi All,
>
> Any input on this would be highly appreciated.
>
> Thanks,
> Vahram.
>
> From: Vahram Aharonyan <va...@vmware.com>
> Sent: Friday, July 6, 2018 4:31 PM
> To: user@geode.apache.org
> Subject: Time difference between cluster members
>
> Hi All,
>
> Could someone help to understand whether there is some threshold on time difference between Geode cluster members if they have the same time zone? Or all members should be in sync from time point of view for normal functioning?
>
> Thanks,
> Vahram.
RE: Time difference between cluster members
Posted by Vahram Aharonyan <va...@vmware.com>.
Hi All,
Any input on this would be highly appreciated.
Thanks,
Vahram.
From: Vahram Aharonyan <va...@vmware.com>
Sent: Friday, July 6, 2018 4:31 PM
To: user@geode.apache.org
Subject: Time difference between cluster members
Hi All,
Could someone help to understand whether there is some threshold on time difference between Geode cluster members if they have the same time zone? Or all members should be in sync from time point of view for normal functioning?
Thanks,
Vahram.