Re: [h2] Data loss with H2 HA / Clustered Mode
No, we have not enabled MVCC so I assume it is disabled. Dan On Saturday, 7 September 2013 05:22:41 UTC+1, Kartweel wrote: On 6/09/2013 11:48 PM, Daniel Stone wrote: We did also experience many problems with a �missing lob� error which actually triggered the cluster issue in the first place.� We have managed to work around this but is this still a bug with H2 (I have seen this problem mentioned on the web)? Are you using MVCC mode by any chance? -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
Hi Stephan, We will take a look at this. Many thanks, Dan On Sunday, 8 September 2013 19:04:54 UTC+1, shess...@googlemail.com wrote: I have a working HA implementation based on H2 which operates on the file system level. It uses H2's file system abstraction and implements a file system class that provides asynchronous replication. If you are interested you may hava a look at https://github.com/shesse/h2ha It's currently based on an older version of H2 (1.3.158) but - as it uses H2 unchanged - it should be not much work to upgrade to a newer H2. Stephan -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
I have a working HA implementation based on H2 which operates on the file system level. It uses H2's file system abstraction and implements a file system class that provides asynchronous replication. If you are interested you may hava a look at https://github.com/shesse/h2ha It's currently based on an older version of H2 (1.3.158) but - as it uses H2 unchanged - it should be not much work to upgrade to a newer H2. Stephan -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
Hi Noel, I hope you don't mind me asking but what is your role within the H2 project? Regards, Dan On Thursday, 5 September 2013 16:17:08 UTC+1, Noel Grandin wrote: On 2013-09-05 17:15, Daniel Stone wrote: Many thanks for the update. We are in the process of deciding whether to stay with H2 or not. I have a few questions regarding H2 clustering. Do you know if there is any intention to enhance the H2 cluster mode? Would we be able to enhance this feature and submit back to the H2 community? I'm not aware of anyone else currently working on the cluster stuff, and we'd be very happy to accept any contributions to this feature! -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
Hi The primary author and project leader is Thomas Mueller. I'm a committer on this project, so I can't speak with final authority, but generally Thomas will speak up if I start spouting nonsense :-) Regards, Noel. On 2013-09-06 11:00, Daniel Stone wrote: Hi Noel, I hope you don't mind me asking but what is your role within the H2 project? Regards, Dan On Thursday, 5 September 2013 16:17:08 UTC+1, Noel Grandin wrote: On 2013-09-05 17:15, Daniel Stone wrote: Many thanks for the update. We are in the process of deciding whether to stay with H2 or not. I have a few questions regarding H2 clustering. Do you know if there is any intention to enhance the H2 cluster mode? Would we be able to enhance this feature and submit back to the H2 community? I'm not aware of anyone else currently working on the cluster stuff, and we'd be very happy to accept any contributions to this feature! -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out. -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
Hi, Thanks for the quick response. We are still in discussions here as to whether we will continue to use H2 with some enhancements etc but I need to establish how complex the process would be if we do. I notice in the FAQ that you recommend patches / bug fixes before making any major changes to the source code, which I can fully understand. How would we go about proposing design / implementation changes to the H2 clustering code? Would anyone with the authority be able to work with us for a donation to the project to speed this process up? Regards, Dan -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
On 2013-09-06 13:01, Daniel Stone wrote: I notice in the FAQ that you recommend patches / bug fixes before making any major changes to the source code, which I can fully understand. How would we go about proposing design / implementation changes to the H2 clustering code? This mailing list is the right place to discuss any proposed changes. Start with a brief breakdown of your requirements i.e. what you see as broken or inadequate. Please don't just post a proposed design. Without knowing what problems you are trying to solve, we won't be able to judge the fitness of any proposed change. Would anyone with the authority be able to work with us for a donation to the project to speed this process up? I don't know of anyone off-hand. Myself and Thomas both have full time jobs and family responsibilities. Fair warning - we're not the fastest moving bunch around :-) Sometimes it can take a couple of days before someone gets around to replying, and large patches can take a week or more before someone finds time to sit down and review it. There is some commercial support available: http://h2database.com/html/links.html#commercial_support -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
Hi, I wrote the current cluster implementation, but I understand it has many limitations. If I would do it again, I would now use a different approach, more like what MongoDB does: use replication for failover and read scalability, and use partitioning for write scalability. The current implementation is basically replication, but it is on the SQL level, which is problematic. I think replication would be better done on a lower level, for example on the file system or table level. But that means the current solution would have to be re-written almost from scratch. Unfortunately, I will not have time to do that in the near future. My priority is (besides fixing bugs of course, and support) to work in the new storage engine (MVStore). Regards, Thomas On Fri, Sep 6, 2013 at 1:13 PM, Noel Grandin noelgran...@gmail.com wrote: On 2013-09-06 13:01, Daniel Stone wrote: I notice in the FAQ that you recommend patches / bug fixes before making any major changes to the source code, which I can fully understand. How would we go about proposing design / implementation changes to the H2 clustering code? This mailing list is the right place to discuss any proposed changes. Start with a brief breakdown of your requirements i.e. what you see as broken or inadequate. Please don't just post a proposed design. Without knowing what problems you are trying to solve, we won't be able to judge the fitness of any proposed change. Would anyone with the authority be able to work with us for a donation to the project to speed this process up? ** ** I don't know of anyone off-hand. Myself and Thomas both have full time jobs and family responsibilities. Fair warning - we're not the fastest moving bunch around :-) Sometimes it can take a couple of days before someone gets around to replying, and large patches can take a week or more before someone finds time to sit down and review it. There is some commercial support available: http://h2database.com/html/links.html#commercial_support -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out. -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
Hi Neol Thomas, Thank you for the responses. I now have enough information to discuss this with my management. I will most likely post my requirements / problems with the current clustering solution next week, possibly with some ideas on how we believe it could be improved. We are very pleased with the performance and general DB functionality of H2 but just need the clustering to handle transient problems more gracefully than it does today. We did also experience many problems with a “missing lob” error which actually triggered the cluster issue in the first place. We have managed to work around this but is this still a bug with H2 (I have seen this problem mentioned on the web)? Regards, Dan On Friday, 6 September 2013 16:22:45 UTC+1, Thomas Mueller wrote: Hi, I wrote the current cluster implementation, but I understand it has many limitations. If I would do it again, I would now use a different approach, more like what MongoDB does: use replication for failover and read scalability, and use partitioning for write scalability. The current implementation is basically replication, but it is on the SQL level, which is problematic. I think replication would be better done on a lower level, for example on the file system or table level. But that means the current solution would have to be re-written almost from scratch. Unfortunately, I will not have time to do that in the near future. My priority is (besides fixing bugs of course, and support) to work in the new storage engine (MVStore). Regards, Thomas On Fri, Sep 6, 2013 at 1:13 PM, Noel Grandin noelg...@gmail.comjavascript: wrote: On 2013-09-06 13:01, Daniel Stone wrote: I notice in the FAQ that you recommend patches / bug fixes before making any major changes to the source code, which I can fully understand. How would we go about proposing design / implementation changes to the H2 clustering code? This mailing list is the right place to discuss any proposed changes. Start with a brief breakdown of your requirements i.e. what you see as broken or inadequate. Please don't just post a proposed design. Without knowing what problems you are trying to solve, we won't be able to judge the fitness of any proposed change. Would anyone with the authority be able to work with us for a donation to the project to speed this process up? ** ** I don't know of anyone off-hand. Myself and Thomas both have full time jobs and family responsibilities. Fair warning - we're not the fastest moving bunch around :-) Sometimes it can take a couple of days before someone gets around to replying, and large patches can take a week or more before someone finds time to sit down and review it. There is some commercial support available: http://h2database.com/html/links.html#commercial_support -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database...@googlegroups.com javascript:. To post to this group, send email to h2-da...@googlegroups.comjavascript: . Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out. -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
On 6/09/2013 11:48 PM, Daniel Stone wrote: We did also experience many problems with a “missing lob” error which actually triggered the cluster issue in the first place. We have managed to work around this but is this still a bug with H2 (I have seen this problem mentioned on the web)? Are you using MVCC mode by any chance? -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
On 2013-09-05 17:15, Daniel Stone wrote: Many thanks for the update. We are in the process of deciding whether to stay with H2 or not. I have a few questions regarding H2 clustering. Do you know if there is any intention to enhance the H2 cluster mode? Would we be able to enhance this feature and submit back to the H2 community? I'm not aware of anyone else currently working on the cluster stuff, and we'd be very happy to accept any contributions to this feature! -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
Hi Noel, Many thanks for the update. We are in the process of deciding whether to stay with H2 or not. I have a few questions regarding H2 clustering. Do you know if there is any intention to enhance the H2 cluster mode? Would we be able to enhance this feature and submit back to the H2 community? Regards, Dan -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
HI On 2013-08-22 19:45, Daniel Stone wrote: My connection URL is essentially this: jdbc:h2:tcp://server1:9092,server2:9092/./folder/dbName OK, so you're not exploring the weirder options, which reduces the problem space considerably. Please note that 2 applications are accessing the cluster. I have now started to look at the H2 source code and from what I can see our system is experiencing transient problems when running updates on server1 which then results in the updates only making it to server2. I then think that another connection (as we use a pool) attempts to read the data from server1 but it doesn't exist. It looks like the cluster list is held per connection, is this correct? Yes, your analysis looks pretty correct. Yes, I suspect that transient failures on one server would cause data out of sync issues. Regards, Noel. -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
Don't think I can help much because I don't use clustering, but let's rule out the obvious problems anyhow :-) - what does your DB URL look like? - are you setting any other DB-level options via SET commands or environment variables? -- Noel. On 2013-08-19 13:26, Daniel Stone wrote: Data loss with H2 HA / Clustered Mode -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
Hi Noel, My connection URL is essentially this: jdbc:h2:tcp://server1:9092,server2:9092/./folder/dbName I have now added ;TRACE_LEVEL_FILE=4 to the URL to switch to SLF4J logging. We do not use any other DB options via SET commands or environment variables. Please note that 2 applications are accessing the cluster. I have now started to look at the H2 source code and from what I can see our system is experiencing transient problems when running updates on server1 which then results in the updates only making it to server2. I then think that another connection (as we use a pool) attempts to read the data from server1 but it doesn't exist. It looks like the cluster list is held per connection, is this correct? Regards, Dan On Thursday, 22 August 2013 16:07:20 UTC+1, Noel Grandin wrote: Don't think I can help much because I don't use clustering, but let's rule out the obvious problems anyhow :-) �- what does your DB URL look like? -� are you setting any other DB-level options via SET commands or environment variables? -- Noel. On 2013-08-19 13:26, Daniel Stone wrote: Data loss with H2 HA / Clustered Mode -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
Re: [h2] Data loss with H2 HA / Clustered Mode
Hi Noel, My connection URL is essentially this: jdbc:h2:tcp://server1:9092,server2:9092/./folder/dbName I have now added ;TRACE_LEVEL_FILE=4 to the URL to switch to SLF4J logging. We do not use any other DB options via SET commands or environment variables. Please note that 2 applications are accessing the cluster. I have now started to look at the H2 source code and from what I can see our system is experiencing transient problems when running updates on server1 which then results in the updates only making it to server2. I then think that another connection (as we use a pool) attempts to read the data from server1 but it doesn't exist. It looks like the cluster list is held per connection, is this correct? Regards, Dan On Thursday, 22 August 2013 16:07:20 UTC+1, Noel Grandin wrote: Don't think I can help much because I don't use clustering, but let's rule out the obvious problems anyhow :-) �- what does your DB URL look like? -� are you setting any other DB-level options via SET commands or environment variables? -- Noel. On 2013-08-19 13:26, Daniel Stone wrote: Data loss with H2 HA / Clustered Mode -- You received this message because you are subscribed to the Google Groups H2 Database group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.