Clusters

[Brett Connor]

We trying clusters with Jackrabbit 1.3.3, and I have a couple of
questions / issues:
 
1) journal database connection recovery - if / when the database is
offline the journal connection is never recovered, and errors are
repeatedly logged. Document operations seem to continue - errors are
logged but no exceptions are thrown. Nodes in the cluster are presumably
out of sync now. Does anyone know if this is recovered cleanly next time
the nodes restart or whether this is permanently bad? I don't like even
asking the question, but I'm being pressured to hold off upgrading to
1.4 at present.
 
2) searching doesn't seem to work over clusters, eg. add a document in
NODE-1, search for the document in NODE-1 works, search with the same
query in NODE-2 and the document is not found. The lucene indexes seem
to be in the local file system always, I can't find any other
configuration. Is there a way for the clusters to either share the
indexes or for JR to update the indexes from the journalled changes?
Does 1.4 change this (I didn't see anything in the release notes about
it)?
 
Thanks
Brett

E-MAIL DISCLAIMER
The information in this e-mail and any attachment is confidential. It is intended only for the named recipient(s). If you are not a named recipient please notify the sender immediately and do not disclose the contents to another person or take copies. Although Axxia Systems has taken every reasonable precaution to ensure that any attachment to this e-mail has been checked for viruses,it is strongly recommended that you carry out your own virus check before opening any attachment, as we cannot accept liability for any damage sustained as a result of software virus infection. Axxia Systems reserves the right to monitor and record e-mails sent to axxia.com, and senders of such messages shall be taken to consent to this.
Registered Office: Axxia House, Unit 4, The Pavilions, Ruscombe Business Park, Twyford, Berkshire, RG10 9NN. Registration Number: 3019229

[Alessandro Bologna]

Regarding point 2, I can tell you that searching *does* work across
clusters. When a clustered server adds, deletes or updates a JCR node,
it writes a record in the journal table and the other servers are
notified and they update their local lucene indexes. Those indexes
need not to be on a shared filesystem.

Sometimes we experimented a situation where the local indexes (stored
on the filesystem, under repository/workspaces/<name>/index) on a
server may have become corrupted, and that leads to queries returning
no nodes. In those cases, it was enough to stop the server that fails,
delete the contents of the index directory, and restart it. Indexes
are regenerated, and it picks up any new change that has happened
while it was down.
Do you see anything weird in your logs?

Alessandro




On Jan 16, 2008 10:17 AM, Connor Brett <ConnorB@...> wrote:

> We trying clusters with Jackrabbit 1.3.3, and I have a couple of
> questions / issues:
>
> 1) journal database connection recovery - if / when the database is
> offline the journal connection is never recovered, and errors are
> repeatedly logged. Document operations seem to continue - errors are
> logged but no exceptions are thrown. Nodes in the cluster are presumably
> out of sync now. Does anyone know if this is recovered cleanly next time
> the nodes restart or whether this is permanently bad? I don't like even
> asking the question, but I'm being pressured to hold off upgrading to
> 1.4 at present.
>
> 2) searching doesn't seem to work over clusters, eg. add a document in
> NODE-1, search for the document in NODE-1 works, search with the same
> query in NODE-2 and the document is not found. The lucene indexes seem
> to be in the local file system always, I can't find any other
> configuration. Is there a way for the clusters to either share the
> indexes or for JR to update the indexes from the journalled changes?
> Does 1.4 change this (I didn't see anything in the release notes about
> it)?
>
> Thanks
> Brett
>
> E-MAIL DISCLAIMER
> The information in this e-mail and any attachment is confidential. It is intended only for the named recipient(s). If you are not a named recipient please notify the sender immediately and do not disclose the contents to another person or take copies. Although Axxia Systems has taken every reasonable precaution to ensure that any attachment to this e-mail has been checked for viruses,it is strongly recommended that you carry out your own virus check before opening any attachment, as we cannot accept liability for any damage sustained as a result of software virus infection. Axxia Systems reserves the right to monitor and record e-mails sent to axxia.com, and senders of such messages shall be taken to consent to this.
> Registered Office: Axxia House, Unit 4, The Pavilions, Ruscombe Business Park, Twyford, Berkshire, RG10 9NN. Registration Number: 3019229
>
>

[Brett Connor]

Thanks, that's good news. What Jackrabbit version you are using? I don't
see anything that stands out in our logs, no errors or warnings anyway.
I haven't examined the entire log to look for the presense of
re-indexing on other nodes. Is there anything I should be looking for,
or any extra configuration I need to do for this to work?

There's no evidence of index corruption - each cluster node can
successfully find document nodes that _it_ added. We have in the past
had the need to drop and re-create the indexes. In my test environment I
started with a brand new repository so all cluster nodes have created
their own local file system indexes.

Thanks for your help.
Brett
 

-----Original Message-----
From: Alessandro Bologna [mailto:alessandro.bologna@...]
Sent: 16 January 2008 16:49
To: users@...
Subject: Re: Clusters

Regarding point 2, I can tell you that searching *does* work across
clusters. When a clustered server adds, deletes or updates a JCR node,
it writes a record in the journal table and the other servers are
notified and they update their local lucene indexes. Those indexes need
not to be on a shared filesystem.

Sometimes we experimented a situation where the local indexes (stored on
the filesystem, under repository/workspaces/<name>/index) on a server
may have become corrupted, and that leads to queries returning no nodes.
In those cases, it was enough to stop the server that fails, delete the
contents of the index directory, and restart it. Indexes are
regenerated, and it picks up any new change that has happened while it
was down.
Do you see anything weird in your logs?

Alessandro




On Jan 16, 2008 10:17 AM, Connor Brett <ConnorB@...> wrote:

> We trying clusters with Jackrabbit 1.3.3, and I have a couple of
> questions / issues:
>
> 1) journal database connection recovery - if / when the database is
> offline the journal connection is never recovered, and errors are
> repeatedly logged. Document operations seem to continue - errors are
> logged but no exceptions are thrown. Nodes in the cluster are
> presumably out of sync now. Does anyone know if this is recovered
> cleanly next time the nodes restart or whether this is permanently
> bad? I don't like even asking the question, but I'm being pressured to

> hold off upgrading to
> 1.4 at present.
>
> 2) searching doesn't seem to work over clusters, eg. add a document in

> NODE-1, search for the document in NODE-1 works, search with the same
> query in NODE-2 and the document is not found. The lucene indexes seem

> to be in the local file system always, I can't find any other
> configuration. Is there a way for the clusters to either share the
> indexes or for JR to update the indexes from the journalled changes?
> Does 1.4 change this (I didn't see anything in the release notes about

> it)?
>
> Thanks
> Brett

E-MAIL DISCLAIMER
The information in this e-mail and any attachment is confidential. It is intended only for the named recipient(s). If you are not a named recipient please notify the sender immediately and do not disclose the contents to another person or take copies. Although Axxia Systems has taken every reasonable precaution to ensure that any attachment to this e-mail has been checked for viruses,it is strongly recommended that you carry out your own virus check before opening any attachment, as we cannot accept liability for any damage sustained as a result of software virus infection. Axxia Systems reserves the right to monitor and record e-mails sent to axxia.com, and senders of such messages shall be taken to consent to this.
Registered Office: Axxia House, Unit 4, The Pavilions, Ruscombe Business Park, Twyford, Berkshire, RG10 9NN. Registration Number: 3019229

[Brett Connor]

I'm embarrassed to say my search problem is solved - I forgot to edit
the node name in the cluster config!


-----Original Message-----
From: Connor Brett [mailto:ConnorB@...]
Sent: 16 January 2008 17:01
To: users@...
Subject: RE: Clusters

Thanks, that's good news. What Jackrabbit version you are using? I don't
see anything that stands out in our logs, no errors or warnings anyway.
I haven't examined the entire log to look for the presense of
re-indexing on other nodes. Is there anything I should be looking for,
or any extra configuration I need to do for this to work?

There's no evidence of index corruption - each cluster node can
successfully find document nodes that _it_ added. We have in the past
had the need to drop and re-create the indexes. In my test environment I
started with a brand new repository so all cluster nodes have created
their own local file system indexes.

Thanks for your help.
Brett
 

-----Original Message-----
From: Alessandro Bologna [mailto:alessandro.bologna@...]
Sent: 16 January 2008 16:49
To: users@...
Subject: Re: Clusters

Regarding point 2, I can tell you that searching *does* work across
clusters. When a clustered server adds, deletes or updates a JCR node,
it writes a record in the journal table and the other servers are
notified and they update their local lucene indexes. Those indexes need
not to be on a shared filesystem.

Sometimes we experimented a situation where the local indexes (stored on
the filesystem, under repository/workspaces/<name>/index) on a server
may have become corrupted, and that leads to queries returning no nodes.
In those cases, it was enough to stop the server that fails, delete the
contents of the index directory, and restart it. Indexes are
regenerated, and it picks up any new change that has happened while it
was down.
Do you see anything weird in your logs?

Alessandro




On Jan 16, 2008 10:17 AM, Connor Brett <ConnorB@...> wrote:

> We trying clusters with Jackrabbit 1.3.3, and I have a couple of
> questions / issues:
>
> 1) journal database connection recovery - if / when the database is
> offline the journal connection is never recovered, and errors are
> repeatedly logged. Document operations seem to continue - errors are
> logged but no exceptions are thrown. Nodes in the cluster are
> presumably out of sync now. Does anyone know if this is recovered
> cleanly next time the nodes restart or whether this is permanently
> bad? I don't like even asking the question, but I'm being pressured to

> hold off upgrading to
> 1.4 at present.
>
> 2) searching doesn't seem to work over clusters, eg. add a document in

> NODE-1, search for the document in NODE-1 works, search with the same
> query in NODE-2 and the document is not found. The lucene indexes seem

> to be in the local file system always, I can't find any other
> configuration. Is there a way for the clusters to either share the
> indexes or for JR to update the indexes from the journalled changes?
> Does 1.4 change this (I didn't see anything in the release notes about

> it)?
>
> Thanks
> Brett

E-MAIL DISCLAIMER
The information in this e-mail and any attachment is confidential. It is intended only for the named recipient(s). If you are not a named recipient please notify the sender immediately and do not disclose the contents to another person or take copies. Although Axxia Systems has taken every reasonable precaution to ensure that any attachment to this e-mail has been checked for viruses,it is strongly recommended that you carry out your own virus check before opening any attachment, as we cannot accept liability for any damage sustained as a result of software virus infection. Axxia Systems reserves the right to monitor and record e-mails sent to axxia.com, and senders of such messages shall be taken to consent to this.
Registered Office: Axxia House, Unit 4, The Pavilions, Ruscombe Business Park, Twyford, Berkshire, RG10 9NN. Registration Number: 3019229

[Stefan Guggisberg]

On Jan 16, 2008 7:33 PM, Connor Brett <ConnorB@...> wrote:
> I'm embarrassed to say my search problem is solved - I forgot to edit
> the node name in the cluster config!

no problem, thanks for reporting back! :)

cheers
stefan

>
>
> -----Original Message-----
> From: Connor Brett [mailto:ConnorB@...]
> Sent: 16 January 2008 17:01
> To: users@...
>
> Subject: RE: Clusters
>
> Thanks, that's good news. What Jackrabbit version you are using? I don't
> see anything that stands out in our logs, no errors or warnings anyway.
> I haven't examined the entire log to look for the presense of
> re-indexing on other nodes. Is there anything I should be looking for,
> or any extra configuration I need to do for this to work?
>
> There's no evidence of index corruption - each cluster node can
> successfully find document nodes that _it_ added. We have in the past
> had the need to drop and re-create the indexes. In my test environment I
> started with a brand new repository so all cluster nodes have created
> their own local file system indexes.
>
> Thanks for your help.
> Brett
>
>
> -----Original Message-----
> From: Alessandro Bologna [mailto:alessandro.bologna@...]
> Sent: 16 January 2008 16:49
> To: users@...
> Subject: Re: Clusters
>
> Regarding point 2, I can tell you that searching *does* work across
> clusters. When a clustered server adds, deletes or updates a JCR node,
> it writes a record in the journal table and the other servers are
> notified and they update their local lucene indexes. Those indexes need
> not to be on a shared filesystem.
>
> Sometimes we experimented a situation where the local indexes (stored on
> the filesystem, under repository/workspaces/<name>/index) on a server
> may have become corrupted, and that leads to queries returning no nodes.
> In those cases, it was enough to stop the server that fails, delete the
> contents of the index directory, and restart it. Indexes are
> regenerated, and it picks up any new change that has happened while it
> was down.
> Do you see anything weird in your logs?
>
> Alessandro
>
>
>
>
> On Jan 16, 2008 10:17 AM, Connor Brett <ConnorB@...> wrote:
> > We trying clusters with Jackrabbit 1.3.3, and I have a couple of
> > questions / issues:
> >
> > 1) journal database connection recovery - if / when the database is
> > offline the journal connection is never recovered, and errors are
> > repeatedly logged. Document operations seem to continue - errors are
> > logged but no exceptions are thrown. Nodes in the cluster are
> > presumably out of sync now. Does anyone know if this is recovered
> > cleanly next time the nodes restart or whether this is permanently
> > bad? I don't like even asking the question, but I'm being pressured to
>
> > hold off upgrading to
> > 1.4 at present.
> >
> > 2) searching doesn't seem to work over clusters, eg. add a document in
>
> > NODE-1, search for the document in NODE-1 works, search with the same
> > query in NODE-2 and the document is not found. The lucene indexes seem
>
> > to be in the local file system always, I can't find any other
> > configuration. Is there a way for the clusters to either share the
> > indexes or for JR to update the indexes from the journalled changes?
> > Does 1.4 change this (I didn't see anything in the release notes about
>
> > it)?
> >
> > Thanks
> > Brett
>
> E-MAIL DISCLAIMER
> The information in this e-mail and any attachment is confidential. It is intended only for the named recipient(s). If you are not a named recipient please notify the sender immediately and do not disclose the contents to another person or take copies. Although Axxia Systems has taken every reasonable precaution to ensure that any attachment to this e-mail has been checked for viruses,it is strongly recommended that you carry out your own virus check before opening any attachment, as we cannot accept liability for any damage sustained as a result of software virus infection. Axxia Systems reserves the right to monitor and record e-mails sent to axxia.com, and senders of such messages shall be taken to consent to this.
> Registered Office: Axxia House, Unit 4, The Pavilions, Ruscombe Business Park, Twyford, Berkshire, RG10 9NN. Registration Number: 3019229
>
>

[Perry.yin]

Hi,stefan:
  I got a jackrabbit clustering issue: Content created by one tomcat is not visible for another tomcat. Would you like to give me your configuration as a sample?
  Thanks in advance!

Stefan Guggisberg wrote:

On Jan 16, 2008 7:33 PM, Connor Brett <[e-mail]> wrote:
> I'm embarrassed to say my search problem is solved - I forgot to edit
> the node name in the cluster config!

no problem, thanks for reporting back! :)

cheers
stefan

>
>
> -----Original Message-----
> From: Connor Brett [mailto:[e-mail]]
> Sent: 16 January 2008 17:01
> To: [e-mail]
>
> Subject: RE: Clusters
>
> Thanks, that's good news. What Jackrabbit version you are using? I don't
> see anything that stands out in our logs, no errors or warnings anyway.
> I haven't examined the entire log to look for the presense of
> re-indexing on other nodes. Is there anything I should be looking for,
> or any extra configuration I need to do for this to work?
>
> There's no evidence of index corruption - each cluster node can
> successfully find document nodes that _it_ added. We have in the past
> had the need to drop and re-create the indexes. In my test environment I
> started with a brand new repository so all cluster nodes have created
> their own local file system indexes.
>
> Thanks for your help.
> Brett
>
>
> -----Original Message-----
> From: Alessandro Bologna [mailto:[e-mail]]
> Sent: 16 January 2008 16:49
> To: [e-mail]
> Subject: Re: Clusters
>
> Regarding point 2, I can tell you that searching *does* work across
> clusters. When a clustered server adds, deletes or updates a JCR node,
> it writes a record in the journal table and the other servers are
> notified and they update their local lucene indexes. Those indexes need
> not to be on a shared filesystem.
>
> Sometimes we experimented a situation where the local indexes (stored on
> the filesystem, under repository/workspaces/<name>/index) on a server
> may have become corrupted, and that leads to queries returning no nodes.
> In those cases, it was enough to stop the server that fails, delete the
> contents of the index directory, and restart it. Indexes are
> regenerated, and it picks up any new change that has happened while it
> was down.
> Do you see anything weird in your logs?
>
> Alessandro
>
>
>
>
> On Jan 16, 2008 10:17 AM, Connor Brett <[e-mail]> wrote:
> > We trying clusters with Jackrabbit 1.3.3, and I have a couple of
> > questions / issues:
> >
> > 1) journal database connection recovery - if / when the database is
> > offline the journal connection is never recovered, and errors are
> > repeatedly logged. Document operations seem to continue - errors are
> > logged but no exceptions are thrown. Nodes in the cluster are
> > presumably out of sync now. Does anyone know if this is recovered
> > cleanly next time the nodes restart or whether this is permanently
> > bad? I don't like even asking the question, but I'm being pressured to
>
> > hold off upgrading to
> > 1.4 at present.
> >
> > 2) searching doesn't seem to work over clusters, eg. add a document in
>
> > NODE-1, search for the document in NODE-1 works, search with the same
> > query in NODE-2 and the document is not found. The lucene indexes seem
>
> > to be in the local file system always, I can't find any other
> > configuration. Is there a way for the clusters to either share the
> > indexes or for JR to update the indexes from the journalled changes?
> > Does 1.4 change this (I didn't see anything in the release notes about
>
> > it)?
> >
> > Thanks
> > Brett
>
> E-MAIL DISCLAIMER
> The information in this e-mail and any attachment is confidential. It is intended only for the named recipient(s). If you are not a named recipient please notify the sender immediately and do not disclose the contents to another person or take copies. Although Axxia Systems has taken every reasonable precaution to ensure that any attachment to this e-mail has been checked for viruses,it is strongly recommended that you carry out your own virus check before opening any attachment, as we cannot accept liability for any damage sustained as a result of software virus infection. Axxia Systems reserves the right to monitor and record e-mails sent to axxia.com, and senders of such messages shall be taken to consent to this.
> Registered Office: Axxia House, Unit 4, The Pavilions, Ruscombe Business Park, Twyford, Berkshire, RG10 9NN. Registration Number: 3019229
>
>

[Stefan Guggisberg-2]

hi perry

On Fri, May 16, 2008 at 10:31 AM, Perry.yin <perry.yin@...> wrote:
>
>  Hi,stefan:
>   I got a jackrabbit clustering issue: Content created by one tomcat is not
>  visible for another tomcat. Would you like to give me your configuration as
>  a sample?

i don't have a cluster setup, sorry. you can find more information
about configuring
cluster support here: http://wiki.apache.org/jackrabbit/Clustering

cheers
stefan

>   Thanks in advance!
>
>
>
>  Stefan Guggisberg wrote:
>  >
>  > On Jan 16, 2008 7:33 PM, Connor Brett <ConnorB@...> wrote:
>  >> I'm embarrassed to say my search problem is solved - I forgot to edit
>  >> the node name in the cluster config!
>  >
>  > no problem, thanks for reporting back! :)
>  >
>  > cheers
>  > stefan
>  >
>  >>
>  >>
>  >> -----Original Message-----
>  >> From: Connor Brett [mailto:ConnorB@...]
>  >> Sent: 16 January 2008 17:01
>  >> To: users@...
>  >>
>  >> Subject: RE: Clusters
>  >>
>  >> Thanks, that's good news. What Jackrabbit version you are using? I don't
>  >> see anything that stands out in our logs, no errors or warnings anyway.
>  >> I haven't examined the entire log to look for the presense of
>  >> re-indexing on other nodes. Is there anything I should be looking for,
>  >> or any extra configuration I need to do for this to work?
>  >>
>  >> There's no evidence of index corruption - each cluster node can
>  >> successfully find document nodes that _it_ added. We have in the past
>  >> had the need to drop and re-create the indexes. In my test environment I
>  >> started with a brand new repository so all cluster nodes have created
>  >> their own local file system indexes.
>  >>
>  >> Thanks for your help.
>  >> Brett
>  >>
>  >>
>  >> -----Original Message-----
>  >> From: Alessandro Bologna [mailto:alessandro.bologna@...]
>  >> Sent: 16 January 2008 16:49
>  >> To: users@...
>  >> Subject: Re: Clusters
>  >>
>  >> Regarding point 2, I can tell you that searching *does* work across
>  >> clusters. When a clustered server adds, deletes or updates a JCR node,
>  >> it writes a record in the journal table and the other servers are
>  >> notified and they update their local lucene indexes. Those indexes need
>  >> not to be on a shared filesystem.
>  >>
>  >> Sometimes we experimented a situation where the local indexes (stored on
>  >> the filesystem, under repository/workspaces/<name>/index) on a server
>  >> may have become corrupted, and that leads to queries returning no nodes.
>  >> In those cases, it was enough to stop the server that fails, delete the
>  >> contents of the index directory, and restart it. Indexes are
>  >> regenerated, and it picks up any new change that has happened while it
>  >> was down.
>  >> Do you see anything weird in your logs?
>  >>
>  >> Alessandro
>  >>
>  >>
>  >>
>  >>
>  >> On Jan 16, 2008 10:17 AM, Connor Brett <ConnorB@...> wrote:
>  >> > We trying clusters with Jackrabbit 1.3.3, and I have a couple of
>  >> > questions / issues:
>  >> >
>  >> > 1) journal database connection recovery - if / when the database is
>  >> > offline the journal connection is never recovered, and errors are
>  >> > repeatedly logged. Document operations seem to continue - errors are
>  >> > logged but no exceptions are thrown. Nodes in the cluster are
>  >> > presumably out of sync now. Does anyone know if this is recovered
>  >> > cleanly next time the nodes restart or whether this is permanently
>  >> > bad? I don't like even asking the question, but I'm being pressured to
>  >>
>  >> > hold off upgrading to
>  >> > 1.4 at present.
>  >> >
>  >> > 2) searching doesn't seem to work over clusters, eg. add a document in
>  >>
>  >> > NODE-1, search for the document in NODE-1 works, search with the same
>  >> > query in NODE-2 and the document is not found. The lucene indexes seem
>  >>
>  >> > to be in the local file system always, I can't find any other
>  >> > configuration. Is there a way for the clusters to either share the
>  >> > indexes or for JR to update the indexes from the journalled changes?
>  >> > Does 1.4 change this (I didn't see anything in the release notes about
>  >>
>  >> > it)?
>  >> >
>  >> > Thanks
>  >> > Brett
>  >>
>  >> E-MAIL DISCLAIMER
>  >> The information in this e-mail and any attachment is confidential. It is
>  >> intended only for the named recipient(s). If you are not a named
>  >> recipient please notify the sender immediately and do not disclose the
>  >> contents to another person or take copies. Although Axxia Systems has
>  >> taken every reasonable precaution to ensure that any attachment to this
>  >> e-mail has been checked for viruses,it is strongly recommended that you
>  >> carry out your own virus check before opening any attachment, as we
>  >> cannot accept liability for any damage sustained as a result of software
>  >> virus infection. Axxia Systems reserves the right to monitor and record
>  >> e-mails sent to axxia.com, and senders of such messages shall be taken to
>  >> consent to this.
>  >> Registered Office: Axxia House, Unit 4, The Pavilions, Ruscombe Business
>  >> Park, Twyford, Berkshire, RG10 9NN. Registration Number: 3019229
>  >>
>  >>
>  >
>  >
>
>  --
>  View this message in context: http://www.nabble.com/Clusters-tp14884942p17269979.html
>  Sent from the Jackrabbit - Users mailing list archive at Nabble.com.
>
>

[Perry.yin]

Hi:
 Thanks a lot, stefan
  And Brett, Would you like to give your clusterting configuration? Very appreciate for that.
I got the following error message :
 "*INFO * SearchManager: Node no longer available 1a6ed71e-531
4-414e-95a2-7632c52a7c5c, skipped. (SearchManager.java, line 467) "

Cheers
Perry

Stefan Guggisberg-2 wrote:

hi perry

On Fri, May 16, 2008 at 10:31 AM, Perry.yin <[e-mail]> wrote:
>
>  Hi,stefan:
>   I got a jackrabbit clustering issue: Content created by one tomcat is not
>  visible for another tomcat. Would you like to give me your configuration as
>  a sample?

i don't have a cluster setup, sorry. you can find more information
about configuring
cluster support here: http://wiki.apache.org/jackrabbit/Clustering

cheers
stefan

>   Thanks in advance!
>
>
>
>  Stefan Guggisberg wrote:
>  >
>  > On Jan 16, 2008 7:33 PM, Connor Brett <[e-mail]> wrote:
>  >> I'm embarrassed to say my search problem is solved - I forgot to edit
>  >> the node name in the cluster config!
>  >
>  > no problem, thanks for reporting back! :)
>  >
>  > cheers
>  > stefan
>  >
>  >>
>  >>
>  >> -----Original Message-----
>  >> From: Connor Brett [mailto:[e-mail]]
>  >> Sent: 16 January 2008 17:01
>  >> To: [e-mail]
>  >>
>  >> Subject: RE: Clusters
>  >>
>  >> Thanks, that's good news. What Jackrabbit version you are using? I don't
>  >> see anything that stands out in our logs, no errors or warnings anyway.
>  >> I haven't examined the entire log to look for the presense of
>  >> re-indexing on other nodes. Is there anything I should be looking for,
>  >> or any extra configuration I need to do for this to work?
>  >>
>  >> There's no evidence of index corruption - each cluster node can
>  >> successfully find document nodes that _it_ added. We have in the past
>  >> had the need to drop and re-create the indexes. In my test environment I
>  >> started with a brand new repository so all cluster nodes have created
>  >> their own local file system indexes.
>  >>
>  >> Thanks for your help.
>  >> Brett
>  >>
>  >>
>  >> -----Original Message-----
>  >> From: Alessandro Bologna [mailto:[e-mail]]
>  >> Sent: 16 January 2008 16:49
>  >> To: [e-mail]
>  >> Subject: Re: Clusters
>  >>
>  >> Regarding point 2, I can tell you that searching *does* work across
>  >> clusters. When a clustered server adds, deletes or updates a JCR node,
>  >> it writes a record in the journal table and the other servers are
>  >> notified and they update their local lucene indexes. Those indexes need
>  >> not to be on a shared filesystem.
>  >>
>  >> Sometimes we experimented a situation where the local indexes (stored on
>  >> the filesystem, under repository/workspaces/<name>/index) on a server
>  >> may have become corrupted, and that leads to queries returning no nodes.
>  >> In those cases, it was enough to stop the server that fails, delete the
>  >> contents of the index directory, and restart it. Indexes are
>  >> regenerated, and it picks up any new change that has happened while it
>  >> was down.
>  >> Do you see anything weird in your logs?
>  >>
>  >> Alessandro
>  >>
>  >>
>  >>
>  >>
>  >> On Jan 16, 2008 10:17 AM, Connor Brett <[e-mail]> wrote:
>  >> > We trying clusters with Jackrabbit 1.3.3, and I have a couple of
>  >> > questions / issues:
>  >> >
>  >> > 1) journal database connection recovery - if / when the database is
>  >> > offline the journal connection is never recovered, and errors are
>  >> > repeatedly logged. Document operations seem to continue - errors are
>  >> > logged but no exceptions are thrown. Nodes in the cluster are
>  >> > presumably out of sync now. Does anyone know if this is recovered
>  >> > cleanly next time the nodes restart or whether this is permanently
>  >> > bad? I don't like even asking the question, but I'm being pressured to
>  >>
>  >> > hold off upgrading to
>  >> > 1.4 at present.
>  >> >
>  >> > 2) searching doesn't seem to work over clusters, eg. add a document in
>  >>
>  >> > NODE-1, search for the document in NODE-1 works, search with the same
>  >> > query in NODE-2 and the document is not found. The lucene indexes seem
>  >>
>  >> > to be in the local file system always, I can't find any other
>  >> > configuration. Is there a way for the clusters to either share the
>  >> > indexes or for JR to update the indexes from the journalled changes?
>  >> > Does 1.4 change this (I didn't see anything in the release notes about
>  >>
>  >> > it)?
>  >> >
>  >> > Thanks
>  >> > Brett
>  >>
>  >> E-MAIL DISCLAIMER
>  >> The information in this e-mail and any attachment is confidential. It is
>  >> intended only for the named recipient(s). If you are not a named
>  >> recipient please notify the sender immediately and do not disclose the
>  >> contents to another person or take copies. Although Axxia Systems has
>  >> taken every reasonable precaution to ensure that any attachment to this
>  >> e-mail has been checked for viruses,it is strongly recommended that you
>  >> carry out your own virus check before opening any attachment, as we
>  >> cannot accept liability for any damage sustained as a result of software
>  >> virus infection. Axxia Systems reserves the right to monitor and record
>  >> e-mails sent to axxia.com, and senders of such messages shall be taken to
>  >> consent to this.
>  >> Registered Office: Axxia House, Unit 4, The Pavilions, Ruscombe Business
>  >> Park, Twyford, Berkshire, RG10 9NN. Registration Number: 3019229
>  >>
>  >>
>  >
>  >
>
>  --
>  View this message in context: http://www.nabble.com/Clusters-tp14884942p17269979.html
>  Sent from the Jackrabbit - Users mailing list archive at Nabble.com.
>
>