Re: Missing LOB issue
I can't argue with you there, and I can understand that this one is a difficult problem, but from what I have read it's been around a while and still is an open problem. I find it strange that more attention has not been paid to it is all. On Thursday, May 30, 2013 4:30:41 PM UTC+10, Noel Grandin wrote: On 2013-05-30 00:34, Jason Pell wrote: We finally got sick of this issue and migrated to HSQLDB for MVCC and no more missing lob issues. We still use H2, but only for storing configuration stuff that does not need to be updated concurrently. It's a shame more attention has not been shown to this issue, because it was certainly frustrating having to change over to HSQLDB. It's an open-source project, so I guess the problem would be yours for failing to provide a fix. (Nobody on this project, to my knowledge, is being paid to work on it). Now, if you want to provide a test-case for your non-MVCC table lock timeout problem, I could maybe help you there, but the MVCC code is still experimental, so I'm not touching that, especially since we're going to drop it when Thomas' new MVStore engine reaches maturity. -- 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?hl=en. For more options, visit https://groups.google.com/groups/opt_out.
Re: Missing LOB issue
Fixing the H2 source is easier said than done, changing to HSQLDB was actually a few hours of work, granted they were a painful few hours, but when you use JPA and Hibernate, switching DB's is relatively painless. We are still using H2 for a lot of stuff, even in production, we just use HSQLDB to replace Oracle in test and dev as it behaves concurrently much better. We use H2 as a configuration store for lots of stuff that would be overkill to use with Oracle. The row level locking is what we really needed and thus because MVCC is experimental with H2, we had to go with another DB where the MVCC support works well with CLOB. I would love to go back to H2 for everything and will certainly keep an eye on MVSTORE. I do contribute to open source myself - being a committer on Apache CXF, but hacking on H2 to fix this particular issue is way beyond my capability especially considering what little time I have. On Thursday, May 30, 2013 4:48:31 PM UTC+10, Kartweel wrote: I'm using MVCC and the only problems I've come across appear to be with LOBS (that's ok for me, don't really need them for my use case). What I mean is that it may have been less work for you to try and fix the H2 source (Seeing the issues seem to all be with LOBS) rather than change to hsqldb (And there would have been other people benefit from that also). Isn't that meant to be one of the good things about open source? I know, I get it from my programmers all the time (They say there is an issue with the library, so they make their own or work around it, which usually doesn't work very well anyway), I tell them to fix it upstream! that is the whole point! And it is getting lots of attention, the entire file store is being re-written to accommodate for MVCC right from the start!, that is why they haven't been giving attention to maintaining the current experimental MVCC mode (Please correct me if I am wrong) On 30/05/2013 2:30 PM, Noel Grandin wrote: On 2013-05-30 00:34, Jason Pell wrote: We finally got sick of this issue and migrated to HSQLDB for MVCC and no more missing lob issues. We still use H2, but only for storing configuration stuff that does not need to be updated concurrently. It's a shame more attention has not been shown to this issue, because it was certainly frustrating having to change over to HSQLDB. It's an open-source project, so I guess the problem would be yours for failing to provide a fix. (Nobody on this project, to my knowledge, is being paid to work on it). Now, if you want to provide a test-case for your non-MVCC table lock timeout problem, I could maybe help you there, but the MVCC code is still experimental, so I'm not touching that, especially since we're going to drop it when Thomas' new MVStore engine reaches maturity. -- 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?hl=en. For more options, visit https://groups.google.com/groups/opt_out.
Re: Missing LOB issue
We are doing integration testing using HSQLDB, we still do full blown integration test with oracle, but on developers PC's Oracle XE is way too heavy, so we have a local build which uses HSQLDB. This does do stuff that requires reasonable concurrency from the DB and tends to cause timeouts otherwise. We always suffered either from timeouts or when we went to MVCC, the missing lobs issue. On Thursday, May 30, 2013 7:40:57 PM UTC+10, Noel Grandin wrote: On 2013-05-30 10:56, jason...@gmail.com javascript: wrote: The real issue is by default h2 locks entire table so only one thread can perform dml stuff this makes it unsuitable for our situation. And we actually only use h2 in test we use Oracle or sql server in production H2 is pretty fast, even with it's simple locking strategy. What are you doing that requires so much speed that single-threaded access to a table is insufficient? Especially since you're only using it in test. -- 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?hl=en. For more options, visit https://groups.google.com/groups/opt_out.
Re: Missing LOB issue
We finally got sick of this issue and migrated to HSQLDB for MVCC and no more missing lob issues. We still use H2, but only for storing configuration stuff that does not need to be updated concurrently. It's a shame more attention has not been shown to this issue, because it was certainly frustrating having to change over to HSQLDB. On Friday, June 15, 2012 6:08:53 PM UTC+10, Noel Grandin wrote: On 2012-06-15 09:54, Eoin Byrne wrote: It looks like we will have to turn MVCC back on because we are now hitting a large number of table lock timeouts but thats going to reintroduce the DB file growth issue. I know this feature is experimental but I guess my question is how experimental is it? It works fine for some people, but others experience problems. The bottom line is that we're just not sure. The people having problems have never been able to produce us a decent test case, so it's kind of hard to fix. and assuming the growth is down to long lived connections - is there any way to find out what selects / updates are still pending commits from the DB logs? Your current best bet is to instrument the H2 code yourself. I suggest these methods org.h2.engine.Session#commit() org.h2.engine.Session#begin() And then you should be able to track the problem sessions down. -- 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?hl=en. For more options, visit https://groups.google.com/groups/opt_out.
1.3.172 when?
Hi, Due to a bug in 1.3.171 with ranges I downgraded to 1.3.169. I would love to upgrade to 1.3.172 when is this due to be available? Thanks Jason -- 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?hl=en. For more options, visit https://groups.google.com/groups/opt_out.
Re: 1.3.172 when?
awesome that will be great. On Tuesday, May 21, 2013 9:32:34 PM UTC+10, Jason Pell wrote: Hi, Due to a bug in 1.3.171 with ranges I downgraded to 1.3.169. I would love to upgrade to 1.3.172 when is this due to be available? Thanks Jason -- 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?hl=en. For more options, visit https://groups.google.com/groups/opt_out.
Re: H2 issue, select with order by, limit and offset throws IllegalArgumentException
1.3.169 does not seem to have this issue. I reverted back to this version once I discovered the problem On Wednesday, May 1, 2013 12:19:56 AM UTC+10, Noel Grandin wrote: Thanks, this is already fixed in SVN. On 2013-04-30 15:12, Dmitry Pekar wrote: create table t (id number); 2. insert into t (id) values (1); 3. select * from t order by id limit 1 offset 2 -- 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?hl=en. For more options, visit https://groups.google.com/groups/opt_out.
