I have a test case with 2 nodes using BDB, replication factor 2,
required writes 2.
I a key/value into a node 1 StoreClient, then pull down node 1.
I read the key from a node 2 StoreClient, and assert that it's there,
but it's not.
The same test worked in 0.80.1.
Anyone got any clues as to why this is?
On Tue, Jan 17, 2012 at 11:16 AM, Steve <stevej...@gmail.com> wrote: > I have a test case with 2 nodes using BDB, replication factor 2, > required writes 2. > I a key/value into a node 1 StoreClient, then pull down node 1. > I read the key from a node 2 StoreClient, and assert that it's there, > but it's not. > The same test worked in 0.80.1. > Anyone got any clues as to why this is?
> -- > You received this message because you are subscribed to the Google Groups > "project-voldemort" group. > To post to this group, send email to project-voldemort@googlegroups.com. > To unsubscribe from this group, send email to > project-voldemort+unsubscribe@googlegroups.com. > For more options, visit this group at > http://groups.google.com/group/project-voldemort?hl=en.
On Wed, Feb 1, 2012 at 12:42 AM, Carlos Tasada <ctas...@gmail.com> wrote: > I'm having a similar problem using 0.90. In my case I have 3 nodes using BDB > with replication factor 2, preferred writes 2, reads 1.
> When I pull down any of the nodes I get an > InsuficientOperationalNodesException.
> Regards,
> On Tue, Jan 17, 2012 at 11:16 AM, Steve <stevej...@gmail.com> wrote:
>> I have a test case with 2 nodes using BDB, replication factor 2, >> required writes 2. >> I a key/value into a node 1 StoreClient, then pull down node 1. >> I read the key from a node 2 StoreClient, and assert that it's there, >> but it's not. >> The same test worked in 0.80.1. >> Anyone got any clues as to why this is?
>> -- >> You received this message because you are subscribed to the Google Groups >> "project-voldemort" group. >> To post to this group, send email to project-voldemort@googlegroups.com. >> To unsubscribe from this group, send email to >> project-voldemort+unsubscribe@googlegroups.com. >> For more options, visit this group at >> http://groups.google.com/group/project-voldemort?hl=en.
> -- > You received this message because you are subscribed to the Google Groups > "project-voldemort" group. > To post to this group, send email to project-voldemort@googlegroups.com. > To unsubscribe from this group, send email to > project-voldemort+unsubscribe@googlegroups.com. > For more options, visit this group at > http://groups.google.com/group/project-voldemort?hl=en.
> If you want read-repair to occur, you should set preferred-reads to 2 as well.
> Thanks,
> - Alex
> On Wed, Feb 1, 2012 at 12:42 AM, Carlos Tasada <ctas...@gmail.com> wrote:
> > I'm having a similar problem using 0.90. In my case I have 3 nodes using BDB
> > with replication factor 2, preferred writes 2, reads 1.
> > When I pull down any of the nodes I get an
> > InsuficientOperationalNodesException.
> > Regards,
> > On Tue, Jan 17, 2012 at 11:16 AM, Steve <stevej...@gmail.com> wrote:
> >> I have a test case with 2 nodes using BDB, replication factor 2,
> >> required writes 2.
> >> I a key/value into a node 1 StoreClient, then pull down node 1.
> >> I read the key from a node 2 StoreClient, and assert that it's there,
> >> but it's not.
> >> The same test worked in 0.80.1.
> >> Anyone got any clues as to why this is?
> >> --
> >> You received this message because you are subscribed to the Google Groups
> >> "project-voldemort" group.
> >> To post to this group, send email to project-voldemort@googlegroups.com.
> >> To unsubscribe from this group, send email to
> >> project-voldemort+unsubscribe@googlegroups.com.
> >> For more options, visit this group at
> >>http://groups.google.com/group/project-voldemort?hl=en.
> > --
> > You received this message because you are subscribed to the Google Groups
> > "project-voldemort" group.
> > To post to this group, send email to project-voldemort@googlegroups.com.
> > To unsubscribe from this group, send email to
> > project-voldemort+unsubscribe@googlegroups.com.
> > For more options, visit this group at
> >http://groups.google.com/group/project-voldemort?hl=en.
> On Feb 7, 9:50 pm, Alex Feinberg <feinb...@gmail.com> wrote:
> > Can you paste your cluster.xml and stores.xml?
> > If you want read-repair to occur, you should set preferred-reads to 2 as well.
> > Thanks,
> > - Alex
> > On Wed, Feb 1, 2012 at 12:42 AM, Carlos Tasada <ctas...@gmail.com> wrote:
> > > I'm having a similar problem using 0.90. In my case I have 3 nodes using BDB
> > > with replication factor 2, preferred writes 2, reads 1.
> > > When I pull down any of the nodes I get an
> > > InsuficientOperationalNodesException.
> > > Regards,
> > > On Tue, Jan 17, 2012 at 11:16 AM, Steve <stevej...@gmail.com> wrote:
> > >> I have a test case with 2 nodes using BDB, replication factor 2,
> > >> required writes 2.
> > >> I a key/value into a node 1 StoreClient, then pull down node 1.
> > >> I read the key from a node 2 StoreClient, and assert that it's there,
> > >> but it's not.
> > >> The same test worked in 0.80.1.
> > >> Anyone got any clues as to why this is?
> > >> --
> > >> You received this message because you are subscribed to the Google Groups
> > >> "project-voldemort" group.
> > >> To post to this group, send email to project-voldemort@googlegroups.com.
> > >> To unsubscribe from this group, send email to
> > >> project-voldemort+unsubscribe@googlegroups.com.
> > >> For more options, visit this group at
> > >>http://groups.google.com/group/project-voldemort?hl=en.
> > > --
> > > You received this message because you are subscribed to the Google Groups
> > > "project-voldemort" group.
> > > To post to this group, send email to project-voldemort@googlegroups.com.
> > > To unsubscribe from this group, send email to
> > > project-voldemort+unsubscribe@googlegroups.com.
> > > For more options, visit this group at
> > >http://groups.google.com/group/project-voldemort?hl=en.
> > On Feb 7, 9:50 pm, Alex Feinberg <feinb...@gmail.com> wrote:
> > > Can you paste your cluster.xml and stores.xml?
> > > If you want read-repair to occur, you should set preferred-reads to 2 > as well.
> > > Thanks, > > > - Alex
> > > On Wed, Feb 1, 2012 at 12:42 AM, Carlos Tasada <ctas...@gmail.com> > wrote: > > > > I'm having a similar problem using 0.90. In my case I have 3 nodes > using BDB > > > > with replication factor 2, preferred writes 2, reads 1.
> > > > When I pull down any of the nodes I get an > > > > InsuficientOperationalNodesException.
> > > > Regards,
> > > > On Tue, Jan 17, 2012 at 11:16 AM, Steve <stevej...@gmail.com> wrote:
> > > >> I have a test case with 2 nodes using BDB, replication factor 2, > > > >> required writes 2. > > > >> I a key/value into a node 1 StoreClient, then pull down node 1. > > > >> I read the key from a node 2 StoreClient, and assert that it's > there, > > > >> but it's not. > > > >> The same test worked in 0.80.1. > > > >> Anyone got any clues as to why this is?
> > > >> -- > > > >> You received this message because you are subscribed to the Google > Groups > > > >> "project-voldemort" group. > > > >> To post to this group, send email to > project-voldemort@googlegroups.com. > > > >> To unsubscribe from this group, send email to > > > >> project-voldemort+unsubscribe@googlegroups.com. > > > >> For more options, visit this group at > > > >>http://groups.google.com/group/project-voldemort?hl=en.
> > > > -- > > > > You received this message because you are subscribed to the Google > Groups > > > > "project-voldemort" group. > > > > To post to this group, send email to > project-voldemort@googlegroups.com. > > > > To unsubscribe from this group, send email to > > > > project-voldemort+unsubscribe@googlegroups.com. > > > > For more options, visit this group at > > > >http://groups.google.com/group/project-voldemort?hl=en.
> -- > You received this message because you are subscribed to the Google Groups > "project-voldemort" group. > To post to this group, send email to project-voldemort@googlegroups.com. > To unsubscribe from this group, send email to > project-voldemort+unsubscribe@googlegroups.com. > For more options, visit this group at > http://groups.google.com/group/project-voldemort?hl=en.
In your case, you have required writes equal to the replication
factor. So, yes, if you pull down a node, you will fail the operation
(an InsufficientOperationalNodesException will be thrown). That is
expected behavior. Also, if you timeout on a single node you'll have
the same problem. Your required reads/writes should generally be at
least 1 less than your replication factor, otherwise you can never
bounce a node without failing your operations. I suggest bringing
required down to 1 and leaving preferred equal to replication factor.
Brendan
On Feb 22, 12:30 am, Carlos Tasada <ctas...@gmail.com> wrote:
> > > On Feb 7, 9:50 pm, Alex Feinberg <feinb...@gmail.com> wrote:
> > > > Can you paste your cluster.xml and stores.xml?
> > > > If you want read-repair to occur, you should set preferred-reads to 2
> > as well.
> > > > Thanks,
> > > > - Alex
> > > > On Wed, Feb 1, 2012 at 12:42 AM, Carlos Tasada <ctas...@gmail.com>
> > wrote:
> > > > > I'm having a similar problem using 0.90. In my case I have 3 nodes
> > using BDB
> > > > > with replication factor 2, preferred writes 2, reads 1.
> > > > > When I pull down any of the nodes I get an
> > > > > InsuficientOperationalNodesException.
> > > > > Regards,
> > > > > On Tue, Jan 17, 2012 at 11:16 AM, Steve <stevej...@gmail.com> wrote:
> > > > >> I have a test case with 2 nodes using BDB, replication factor 2,
> > > > >> required writes 2.
> > > > >> I a key/value into a node 1 StoreClient, then pull down node 1.
> > > > >> I read the key from a node 2 StoreClient, and assert that it's
> > there,
> > > > >> but it's not.
> > > > >> The same test worked in 0.80.1.
> > > > >> Anyone got any clues as to why this is?
> > > > >> --
> > > > >> You received this message because you are subscribed to the Google
> > Groups
> > > > >> "project-voldemort" group.
> > > > >> To post to this group, send email to
> > project-voldemort@googlegroups.com.
> > > > >> To unsubscribe from this group, send email to
> > > > >> project-voldemort+unsubscribe@googlegroups.com.
> > > > >> For more options, visit this group at
> > > > >>http://groups.google.com/group/project-voldemort?hl=en.
> > > > > --
> > > > > You received this message because you are subscribed to the Google
> > Groups
> > > > > "project-voldemort" group.
> > > > > To post to this group, send email to
> > project-voldemort@googlegroups.com.
> > > > > To unsubscribe from this group, send email to
> > > > > project-voldemort+unsubscribe@googlegroups.com.
> > > > > For more options, visit this group at
> > > > >http://groups.google.com/group/project-voldemort?hl=en.
> > --
> > You received this message because you are subscribed to the Google Groups
> > "project-voldemort" group.
> > To post to this group, send email to project-voldemort@googlegroups.com.
> > To unsubscribe from this group, send email to
> > project-voldemort+unsubscribe@googlegroups.com.
> > For more options, visit this group at
> >http://groups.google.com/group/project-voldemort?hl=en.
On Jan 17, 2:16 am, Steve <stevej...@gmail.com> wrote:
> I have a test case with 2 nodes using BDB, replication factor 2,
> required writes 2.
> I a key/value into a node 1 StoreClient, then pull down node 1.
> I read the key from a node 2 StoreClient, and assert that it's there,
> but it's not.
> The same test worked in 0.80.1.
> Anyone got any clues as to why this is?
As Alex asked earlier, can we please see your stores.xml and
cluster.xml? We cannot really help you without first seeing that.
So if I'm getting this right, if one node is down, the replication-factor=2 and the preferred-reads=2 will force the key sync when the node is up again?
On Wed, Feb 22, 2012 at 9:57 AM, Brendan Harris <voldem...@stotch.com>wrote:
> On Jan 17, 2:16 am, Steve <stevej...@gmail.com> wrote: > > I have a test case with 2 nodes using BDB, replication factor 2, > > required writes 2. > > I a key/value into a node 1 StoreClient, then pull down node 1. > > I read the key from a node 2 StoreClient, and assert that it's there, > > but it's not. > > The same test worked in 0.80.1. > > Anyone got any clues as to why this is?
> As Alex asked earlier, can we please see your stores.xml and > cluster.xml? We cannot really help you without first seeing that.
> -- > You received this message because you are subscribed to the Google Groups > "project-voldemort" group. > To post to this group, send email to project-voldemort@googlegroups.com. > To unsubscribe from this group, send email to > project-voldemort+unsubscribe@googlegroups.com. > For more options, visit this group at > http://groups.google.com/group/project-voldemort?hl=en.
If you are configured to use hinted handoff, then the data will be pushed to the down node from a peer node automatically when it recovers--that's independent of preferred-reads. Also, with preferred-reads equal to the replication factor as you have it, then read repair would sync up the data upon read of the key, independent of hinted handoff. So there are two independent mechanisms working to fix up the data.
On Wed, Feb 22, 2012 at 2:22 AM, Carlos Tasada <ctas...@gmail.com> wrote: > So if I'm getting this right, if one node is down, the > replication-factor=2 and the preferred-reads=2 will force the key sync when > the node is up again?
On Thursday, March 1, 2012 4:43:31 AM UTC+1, Jeffstyr wrote:
> If you are configured to use hinted handoff, then the data will be pushed > to the down node from a peer node automatically when it recovers--that's > independent of preferred-reads. Also, with preferred-reads equal to the > replication factor as you have it, then read repair would sync up the data > upon read of the key, independent of hinted handoff. So there are two > independent mechanisms working to fix up the data.
> On Wed, Feb 22, 2012 at 2:22 AM, Carlos Tasada <ctas...@gmail.com> wrote:
>> So if I'm getting this right, if one node is down, the >> replication-factor=2 and the preferred-reads=2 will force the key sync when >> the node is up again?
We reworked the upgrade and the problem did not repeat. I didn't keep
the patch for the original failed upgrade, so I'm unsure what I did
wrong.
Thanks to you and Alex for getting back to me though.
Steve
On Feb 22, 4:57 pm, Brendan Harris <voldem...@stotch.com> wrote:
> On Jan 17, 2:16 am, Steve <stevej...@gmail.com> wrote:
> > I have a test case with 2 nodes using BDB, replication factor 2,
> > required writes 2.
> > I a key/value into a node 1 StoreClient, then pull down node 1.
> > I read the key from a node 2 StoreClient, and assert that it's there,
> > but it's not.
> > The same test worked in 0.80.1.
> > Anyone got any clues as to why this is?
> As Alex asked earlier, can we please see your stores.xml and
> cluster.xml? We cannot really help you without first seeing that.