Re: [Bacula-users] Tuning for large (millions of files) backups?
On Thu, Nov 18, 2010 at 10:27 PM, Alan Brown a...@mssl.ucl.ac.uk wrote: On 13/11/10 04:46, Gary R. Schmidt wrote: You mean looks increasingly *unlikely* don't you? As InnoDB is the default in MySQL 5.5... Yes it is, but take a look at what Oracle's been doing to the other opensource projects it inherited. It says a lot when core mysql developers fork a new project. It says a lot more when this happens across a number of projects including the entire Open Office developer team. That's quite an exaggeration, although I am definitely not feeling good about the amount of bad will Oracle managed to inspire in me in this short amount of time. Frankly, the only thing going for them in my book right now IS MySQL 5.5. I suspect there to be at least one person involved in this discussion who has *religion* in relation to database engines... Nothing to do with religion - and FWIW, stating that postgresql requires a DBA is a clear case of FUD. My point of view comes from running both engines on the same hardware and observing the loads involved. Personally I'd prefer to be running mysql but it was clear postgres ran faster and had lower memory foorprints for our use than innodb. Others have reported the same thing over the years. The FUD stops here, this is pointless in the case of (where this discussion started) restore performance on a MySQL back-end. The SQL queries are not at all written with performance for MySQL in mind. And frankly, the file selection process shouldn't even pull the entire file list from the database at once, it should be done kindof like the .Bvfs API is built, with the Hierarchy in mind. Pulling 5 million files out in one flat list is equally stupid to (or, rather in this case, a simplification) storing 5 million files in an unhashed directory structure. As soon as you see subqueries like these run against a MySQL server it is obvious it was not designed for MySQL and/or performance. Frankly I don't know at this point how to make it better without restructuring the database and actually avoiding pulling out millions of millions of records at once. Hierarchies are definitely the way to go in this case, as it was a question of restore selection. Remember, this was about file selection, you don't actually need the full list of million(s) if you're only going to choose a subset. But you do need the ability to traverse. Here's the query which is the foundation of this entire thread, tidied up a bit. Anyone who's ever dealt with MySQL can see that this is not going to look good in EXPLAIN: SELECT Path.Path, Filename.Name, Temp.FileIndex, Temp.JobId, LStat, MD5 FROM ( SELECT FileId, Job.JobId AS JobId, FileIndex, File.PathId AS PathId, File.FilenameId AS FilenameId, LStat, MD5 FROM Job, File, ( SELECT MAX(JobTDate) AS JobTDate, PathId, FilenameId FROM ( SELECT JobTDate, PathId, FilenameId FROM File JOIN Job USING (JobId) WHERE File.JobId IN (38,39) UNION ALL SELECT JobTDate, PathId, FilenameId FROM BaseFiles JOIN File USING (FileId) JOIN Job ON (BaseJobId = Job.JobId) WHERE BaseFiles.JobId IN (38,39) ) AS tmp GROUP BY PathId, FilenameId ) AS T1 WHERE (Job.JobId IN ( SELECT DISTINCT BaseJobId FROM BaseFiles WHERE JobId IN (38,39) ) OR Job.JobId IN (38,39) ) AND T1.JobTDate = Job.JobTDate AND Job.JobId = File.JobId AND T1.PathId = File.PathId AND T1.FilenameId = File.FilenameId ) AS Temp JOIN Filename ON (Filename.FilenameId = Temp.FilenameId) JOIN Path ON (Path.PathId = Temp.PathId) WHERE FileIndex 0 ORDER BY Temp.JobId, FileIndex ASC; -- Mikael -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today http://p.sf.net/sfu/msIE9-sfdev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
On Fri, 19 Nov 2010, Mikael Fridh wrote: The FUD stops here, this is pointless in the case of (where this discussion started) restore performance on a MySQL back-end. In terms of restore performance, you're right. Better optimised queries would speed things up, but probably not by much (see below: Bacula-dir is the major factor on large restores) Pulling 5 million files out in one flat list is equally stupid to (or, rather in this case, a simplification) storing 5 million files in an unhashed directory structure. I _have_ users who do that. (arrgh!) As soon as you see subqueries like these run against a MySQL server it is obvious it was not designed for MySQL and/or performance. In both cases (mysql or postgres) the actual query is relatively fast for 1-2 million file backups, but then bacula-dir itself grinds on the results for quite some time and it's memory-intensive while doing it. I've pointed Kern at alternatives to red/black tables which will probably speed that side up, but optimised queries are always a good idea. Frankly I don't know at this point how to make it better without restructuring the database and actually avoiding pulling out millions of millions of records at once. If you can make a better mousetrap, Kern (and a lot of other people) will probably thank you - even if it takes a major revision release to change the database format. FWIW, the EXPLAIN is just as ugly on postgres. AB -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today http://p.sf.net/sfu/msIE9-sfdev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
On 13/11/10 04:46, Gary R. Schmidt wrote: You mean looks increasingly *unlikely* don't you? As InnoDB is the default in MySQL 5.5... Yes it is, but take a look at what Oracle's been doing to the other opensource projects it inherited. It says a lot when core mysql developers fork a new project. It says a lot more when this happens across a number of projects including the entire Open Office developer team. I suspect there to be at least one person involved in this discussion who has *religion* in relation to database engines... Nothing to do with religion - and FWIW, stating that postgresql requires a DBA is a clear case of FUD. My point of view comes from running both engines on the same hardware and observing the loads involved. Personally I'd prefer to be running mysql but it was clear postgres ran faster and had lower memory foorprints for our use than innodb. Others have reported the same thing over the years. (Longer term I'm concerned about what Oracle may do with Mysql as we have a number of databases installed on various machines machines doing various things for various groups space scientists are difficult to deal with at the best of times, let alone if they have to change tools.) Frankly, I'd rather there were reliable connectors and queries available for Oracle and DB2, rather than this childish prattle over MySQL and PostGRES. It'd be nice, but it's not going to happen unless someone who wants them, writes it (or pays for it) -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today http://p.sf.net/sfu/msIE9-sfdev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
Hi, On Fri, 12 Nov 2010, Bob Hetzel wrote: I'm starting to think the issue might be linked to some kernels or linux distros. I have two bacula servers here. One system is a year and a half old (12 GB RAM), has with a File table having approx 40 million File records. That system has had the slowness issue (building the directory tree on restores took about an hour) running first Ubuntu 9.04 or 9.10 and now RedHat 6 beta. The kernel currently is at 2.6.32-44.1.el6.x86_64. I haven't tried downgrading, instead I tweaked the source code to use the old 3.0.3 query and recompiled--I don't use Base jobs or Accurate backups so that's safe for me. The other system is 4 yrs or so old, with less memory (8GB), slower cpus, slower hard drives, etc., and in fairness only 35 million File records. This one builds the directory tree in approx 10 seconds, but is running Centos 5.5. The kernel currently is at 2.6.18-194.11.3.el5. That's an interesting thought. It would be interesting to make an exact comparison, something like: - run a restore with the slow query log on, capture the query text - run the query manually in mysql - dump the mysql database - restore the mysql database on the older server - run the query there It sounds like your database is quite large so this might be too awkward in practice? Strictly speaking the freshly sequentially written database might have a slight unfair advantage, but if the results are radically different then that would be useful to know. Gavin -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
On 11/12/2010 11:46 PM, Gary R. Schmidt wrote: Frankly, I'd rather there were reliable connectors and queries available for Oracle and DB2 My usual conclusion when something does not exist is that nobody [with the ability to create it] wants them. rather than this childish prattle over MySQL and PostGRES. Anyone can stop it any time. -- Dan Langille - http://langille.org/ -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
On Thu, Nov 11, 2010 at 3:47 PM, Gavin McCullagh gavin.mccull...@gcd.ie wrote: On Mon, 08 Nov 2010, Gavin McCullagh wrote: We seem to have the correct indexes on the file table. I've run optimize table and it still takes 14 minutes to build the tree on one of our bigger clients. We have 51 million entries in the file table. I thought I should give some mroe concrete information: I don't suppose this is news to anyone but here's the mysql slow query log to correspond: # Time: 10 14:24:49 # u...@host: bacula[bacula] @ localhost [] # Query_time: 1139.657646 Lock_time: 0.000471 Rows_sent: 4263403 Rows_examined: 50351037 SET timestamp=1289485489; SELECT Path.Path, Filename.Name, Temp.FileIndex, Temp.JobId, LStat, MD5 FROM ( SELECT FileId, Job.JobId AS JobId, FileIndex, File.PathId AS PathId, File.FilenameId AS FilenameId, LStat, MD5 FROM Job, File, ( SELECT MAX(JobTDate) AS JobTDate, PathId, FilenameId FROM ( SELECT JobTDate, PathId, FilenameId FROM File JOIN Job USING (JobId) WHERE File.JobId IN (9944,9950,9973,9996) UNION ALL SELECT JobTDate, PathId, FilenameId FROM BaseFiles JOIN File USING (FileId) JOIN Job ON (BaseJobId = Job.JobId) WHERE BaseFiles.JobId IN (9944,9950,9973,9996) ) AS tmp GROUP BY PathId, FilenameId ) AS T1 WHERE (Job.JobId IN ( SELECT DISTINCT BaseJobId FROM BaseFiles WHERE JobId IN (9944,9950,9973,9996)) OR Job.JobId IN (9944,9950,9973,9996)) AND T1.JobTDate = Job.JobTDate AND Job.JobId = File.JobId AND T1.PathId = File.PathId AND T1.FilenameId = File.FilenameId ) AS Temp JOIN Filename ON (Filename.FilenameId = Temp.FilenameId) JOIN Path ON (Path.PathId = Temp.PathId) WHERE FileIndex 0 ORDER BY Temp.JobId, FileIndex ASC; Could you please do an EXPLAIN on this query? I know it's going to look awful but I'm curious anyway. subqueries like these and SELECT DISTINCT are usually a recipe for disastrous querytimes in MySQL. I've spent some time with the mysqltuner.pl script but to no avail thus far. There's 6GB RAM so it suggests a key buffer size of 4GB which I've set at 4.1GB. Tuning's not going to get any of those 50 million traversed rows disappear. Only a differently optimized query plan will. This is an Ubuntu Linux server running MySQL v5.1.41. The mysql data is on an MD software RAID 1 array on 7200rpm SATA disks. The tables are MyISAM (which I had understood to be quicker than innodb in low concurrency situations?). The tuner script is suggesting I should disable innodb as we're not using it which I will do though I wouldn't guess that will make a massive difference. No, it will not help. -- Mikael -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
Hi, On Fri, 12 Nov 2010, Mikael Fridh wrote: On Thu, Nov 11, 2010 at 3:47 PM, Gavin McCullagh gavin.mccull...@gcd.ie wrote: # Time: 10 14:24:49 # u...@host: bacula[bacula] @ localhost [] # Query_time: 1139.657646 Lock_time: 0.000471 Rows_sent: 4263403 Rows_examined: 50351037 SET timestamp=1289485489; SELECT Path.Path, Filename.Name, Temp.FileIndex, Temp.JobId, LStat, MD5 FROM ( SELECT FileId, Job.JobId AS JobId, FileIndex, File.PathId AS PathId, File.FilenameId AS FilenameId, LStat, MD5 FROM Job, File, ( SELECT MAX(JobTDate) AS JobTDate, PathId, FilenameId FROM ( SELECT JobTDate, PathId, FilenameId FROM File JOIN Job USING (JobId) WHERE File.JobId IN (9944,9950,9973,9996) UNION ALL SELECT JobTDate, PathId, FilenameId FROM BaseFiles JOIN File USING (FileId) JOIN Job ON (BaseJobId = Job.JobId) WHERE BaseFiles.JobId IN (9944,9950,9973,9996) ) AS tmp GROUP BY PathId, FilenameId ) AS T1 WHERE (Job.JobId IN ( SELECT DISTINCT BaseJobId FROM BaseFiles WHERE JobId IN (9944,9950,9973,9996)) OR Job.JobId IN (9944,9950,9973,9996)) AND T1.JobTDate = Job.JobTDate AND Job.JobId = File.JobId AND T1.PathId = File.PathId AND T1.FilenameId = File.FilenameId ) AS Temp JOIN Filename ON (Filename.FilenameId = Temp.FilenameId) JOIN Path ON (Path.PathId = Temp.PathId) WHERE FileIndex 0 ORDER BY Temp.JobId, FileIndex ASC; Could you please do an EXPLAIN on this query? I prefixed the query by the word EXPLAIN and ran it: mysql source bacularestorequery.sql +++++-++-+-+-+-+ | id | select_type| table | type | possible_keys | key| key_len | ref | rows| Extra | +++++-++-+-+-+-+ | 1 | PRIMARY| derived2 | ALL| NULL | NULL | NULL| NULL| 4277605 | Using where; Using filesort | | 1 | PRIMARY| Filename | eq_ref | PRIMARY | PRIMARY| 4 | Temp.FilenameId | 1 | | | 1 | PRIMARY| Path | eq_ref | PRIMARY | PRIMARY| 4 | Temp.PathId | 1 | | | 2 | DERIVED| derived3 | ALL| NULL | NULL | NULL| NULL| 4277605 | | | 2 | DERIVED| File | ref| PathId,FilenameId,JobId,jobid_index | FilenameId | 8 | T1.FilenameId,T1.PathId | 4 | Using where | | 2 | DERIVED| Job| eq_ref | PRIMARY | PRIMARY| 4 | bacula.File.JobId | 1 | Using where | | 6 | DEPENDENT SUBQUERY | NULL | NULL | NULL | NULL | NULL| NULL|NULL | no matching row in const table | | 3 | DERIVED| derived4 | ALL| NULL | NULL | NULL| NULL| 4302683 | Using temporary; Using filesort | | 4 | DERIVED| Job| range | PRIMARY | PRIMARY| 4 | NULL| 4 | Using where | | 4 | DERIVED| File | ref| JobId,jobid_index | JobId | 4 | bacula.Job.JobId| 41816 | Using index | | 5 | UNION | NULL | NULL | NULL | NULL | NULL| NULL|NULL | no matching row in const table | | NULL | UNION RESULT | union4,5 | ALL| NULL | NULL | NULL| NULL|NULL | | +++++-++-+-+-+-+ 12 rows in set (16 min 15.79 sec) I presume that's what you're looking for? Tuning's not going to get any of those 50 million traversed rows disappear. Only a differently optimized query plan will. Well, if the above helps and/or if you'd like me to run an alternative proposed query I'm happy to. I must confess it would take me quite a few hours to actually understand that query. Gavin -- Centralized Desktop Delivery: Dell and VMware Reference Architecture
Re: [Bacula-users] Tuning for large (millions of files) backups?
Mikael Fridh wrote: Tuning's not going to get any of those 50 million traversed rows disappear. Only a differently optimized query plan will. This applies across both mysql and postgresql... This is an Ubuntu Linux server running MySQL v5.1.41. The mysql data is on an MD software RAID 1 array on 7200rpm SATA disks. The tables are MyISAM (which I had understood to be quicker than innodb in low concurrency situations?). The tuner script is suggesting I should disable innodb as we're not using it which I will do though I wouldn't guess that will make a massive difference. No, it will not help. Disbling innodb won't help right now, but switching to innodb would be a good idea in the near future as myISAM runs into problems around the 50 million entry mark (assuming Oracle don't remove innodb from future versions of Mysql, as looks increasingly likely...) -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
'Alan Brown' wrote: Mikael Fridh wrote: Tuning's not going to get any of those 50 million traversed rows disappear. Only a differently optimized query plan will. This applies across both mysql and postgresql... This is an Ubuntu Linux server running MySQL v5.1.41. The mysql data is on an MD software RAID 1 array on 7200rpm SATA disks. The tables are MyISAM (which I had understood to be quicker than innodb in low concurrency situations?). The tuner script is suggesting I should disable innodb as we're not using it which I will do though I wouldn't guess that will make a massive difference. No, it will not help. Disbling innodb won't help right now, but switching to innodb would be a good idea in the near future as myISAM runs into problems around the 50 million entry mark (assuming Oracle don't remove innodb from future versions of Mysql, as looks increasingly likely...) InnoDB is the default storage engine for MySQL 5.5 -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Med venlig hilsen / Best Regards Henrik Johansen hen...@scannet.dk Tlf. 75 53 35 00 ScanNet Group A/S ScanNet -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
From: Gavin McCullagh gavin.mccull...@gcd.ie Subject: Re: [Bacula-users] Tuning for large (millions of files) backups? To: bacula-users@lists.sourceforge.net Message-ID: 2010144733.gz20...@gcd.ie Content-Type: text/plain; charset=us-ascii On Mon, 08 Nov 2010, Gavin McCullagh wrote: We seem to have the correct indexes on the file table. I've run optimize table and it still takes 14 minutes to build the tree on one of our bigger clients. We have 51 million entries in the file table. I thought I should give some mroe concrete information: I don't suppose this is news to anyone but here's the mysql slow query log to correspond: # Time: 10 14:24:49 # u...@host: bacula[bacula] @ localhost [] # Query_time: 1139.657646 Lock_time: 0.000471 Rows_sent: 4263403 Rows_examined: 50351037 SET timestamp=1289485489; SELECT Path.Path, Filename.Name, Temp.FileIndex, Temp.JobId, LStat, MD5 FROM ( SELECT FileId, Job.JobId AS JobId, FileIndex, File.PathId AS PathId, File.FilenameId AS FilenameId, LStat, MD5 FROM Job, File, ( SELECT MAX(JobTDate) AS JobTDate, PathId, FilenameId FROM ( SELECT JobTDate, PathId, FilenameId FROM File JOIN Job USING (JobId) WHERE File.JobId IN (9944,9950,9973,9996) UNION ALL SELECT JobTDate, PathId, FilenameId FROM BaseFiles JOIN File USING (FileId) JOIN Job ON(BaseJobId = Job.JobId) WHERE BaseFiles.JobId IN (9944,9950,9973,9996) ) AS tmp GROUP BY PathId, FilenameId ) AS T1 WHERE (Job.JobId IN ( SELECT DISTINCT BaseJobId FROM BaseFiles WHERE JobId IN (9944,9950,9973,9996)) OR Job.JobId IN (9944,9950,9973,9996)) AND T1.JobTDate = Job.JobTDate AND Job.JobId = File.JobId AND T1.PathId = File.PathId AND T1.FilenameId = File.FilenameId ) AS Temp JOIN Filename ON (Filename.FilenameId = Temp.FilenameId) JOIN Path ON (Path.PathId = Temp.PathId) WHERE FileIndex 0 ORDE! R BY Temp.JobId, FileIndex ASC; I've spent some time with the mysqltuner.pl script but to no avail thus far. There's 6GB RAM so it suggests a key buffer size of 4GB which I've set at 4.1GB. This is an Ubuntu Linux server running MySQL v5.1.41. The mysql data is on an MD software RAID 1 array on 7200rpm SATA disks. The tables are MyISAM (which I had understood to be quicker than innodb in low concurrency situations?). The tuner script is suggesting I should disable innodb as we're not using it which I will do though I wouldn't guess that will make a massive difference. There are no fragmented tables currently. Gavin I'm starting to think the issue might be linked to some kernels or linux distros. I have two bacula servers here. One system is a year and a half old (12 GB RAM), has with a File table having approx 40 million File records. That system has had the slowness issue (building the directory tree on restores took about an hour) running first Ubuntu 9.04 or 9.10 and now RedHat 6 beta. The kernel currently is at 2.6.32-44.1.el6.x86_64. I haven't tried downgrading, instead I tweaked the source code to use the old 3.0.3 query and recompiled--I don't use Base jobs or Accurate backups so that's safe for me. The other system is 4 yrs or so old, with less memory (8GB), slower cpus, slower hard drives, etc., and in fairness only 35 million File records. This one builds the directory tree in approx 10 seconds, but is running Centos 5.5. The kernel currently is at 2.6.18-194.11.3.el5. I'm still convinced that this one slow MySQL query could be changed to allow MySQL to better optimize it. I started with the same my.cnf file settings and then tried tweaking them because the newer computer has more ram but that didn't help. Is anybody up to the task of rewriting that query? -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
Gavin McCullagh wrote: On Tue, 09 Nov 2010, Alan Brown wrote: and it still takes 14 minutes to build the tree on one of our bigger clients. We have 51 million entries in the file table. Add individual indexes for Fileid, Jobid and Pathid Postgres will work with the combined index for individual table queries, but mysql won't. The following are the indexes on the file table: mysql SHOW INDEXES FROM File; +---++--+--+-+---+-+--++--++-+ | Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | +---++--+--+-+---+-+--++--++-+ | File | 0 | PRIMARY |1 | FileId | A |55861148 | NULL | NULL | | BTREE | | | File | 1 | PathId |1 | PathId | A | 735015 | NULL | NULL | | BTREE | | | File | 1 | FilenameId |1 | FilenameId | A | 2539143 | NULL | NULL | | BTREE | | | File | 1 | FilenameId |2 | PathId | A |13965287 | NULL | NULL | | BTREE | | | File | 1 | JobId|1 | JobId | A |1324 | NULL | NULL | | BTREE | | | File | 1 | JobId|2 | PathId | A | 2940060 | NULL | NULL | | BTREE | | | File | 1 | JobId|3 | FilenameId | A |55861148 | NULL | NULL | | BTREE | | | File | 1 | jobid_index |1 | JobId | A |1324 | NULL | NULL | | BTREE | | | File | 1 | pathid_index |1 | PathId | A | 735015 | NULL | NULL | | BTREE | | +---++--+--+-+---+-+--++--++-+ I added the last two per your instructions. Building the tree took about 14 minutes without these indexes and takes about 17-18 minutes having added them. What tuning (if any) have you performed on your my.cnf and how much memory do you have? Have I done something wrong? As FileId is a primary key, it doesn't seem like I should need an extra index on that one -- is that wrong? It doesn't need an extra index. You've also got a duplicate pathid indeax which can be deleted. This kind of thing is why it makes more sense to switch to postgres when mysql databases get large. -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
'Alan Brown' wrote: Gavin McCullagh wrote: On Tue, 09 Nov 2010, Alan Brown wrote: and it still takes 14 minutes to build the tree on one of our bigger clients. We have 51 million entries in the file table. Add individual indexes for Fileid, Jobid and Pathid Postgres will work with the combined index for individual table queries, but mysql won't. The following are the indexes on the file table: mysql SHOW INDEXES FROM File; +---++--+--+-+---+-+--++--++-+ | Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | +---++--+--+-+---+-+--++--++-+ | File | 0 | PRIMARY |1 | FileId | A |55861148 | NULL | NULL | | BTREE | | | File | 1 | PathId |1 | PathId | A | 735015 | NULL | NULL | | BTREE | | | File | 1 | FilenameId |1 | FilenameId | A | 2539143 | NULL | NULL | | BTREE | | | File | 1 | FilenameId |2 | PathId | A |13965287 | NULL | NULL | | BTREE | | | File | 1 | JobId|1 | JobId | A |1324 | NULL | NULL | | BTREE | | | File | 1 | JobId|2 | PathId | A | 2940060 | NULL | NULL | | BTREE | | | File | 1 | JobId|3 | FilenameId | A |55861148 | NULL | NULL | | BTREE | | | File | 1 | jobid_index |1 | JobId | A |1324 | NULL | NULL | | BTREE | | | File | 1 | pathid_index |1 | PathId | A | 735015 | NULL | NULL | | BTREE | | +---++--+--+-+---+-+--++--++-+ I added the last two per your instructions. Building the tree took about 14 minutes without these indexes and takes about 17-18 minutes having added them. What tuning (if any) have you performed on your my.cnf and how much memory do you have? Have I done something wrong? As FileId is a primary key, it doesn't seem like I should need an extra index on that one -- is that wrong? It doesn't need an extra index. You've also got a duplicate pathid indeax which can be deleted. This kind of thing is why it makes more sense to switch to postgres when mysql databases get large. I have had about as much of this as I can take now so please, stop spreading FUD about MySQL. When it comes to Bacula there is only one valid concern - Postgres has certain statement constructs which allow certain queries to be performed faster - that's about it. It am not buying the postulation that postgres is largely self-tuning, especially not when dealing with large datasets. If you prefer postgres, that's totally fine but please stop telling people that MySQL is unusable for large DB deployments because this simply is untrue. -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Med venlig hilsen / Best Regards Henrik Johansen hen...@scannet.dk Tlf. 75 53 35 00 ScanNet Group A/S ScanNet -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
Hi, On Thu, 11 Nov 2010, Alan Brown wrote: What tuning (if any) have you performed on your my.cnf and how much memory do you have? Thus far I haven't spent much time on this and haven't tuned MySQL. The slow build an annoyance, but not a killer so I've not really got around to it. The server has 6GB RAM (running an X86_64 kernel). Have I done something wrong? As FileId is a primary key, it doesn't seem like I should need an extra index on that one -- is that wrong? It doesn't need an extra index. Grand. You've also got a duplicate pathid indeax which can be deleted. Ah, I didn't spot that, thanks. This kind of thing is why it makes more sense to switch to postgres when mysql databases get large. I see. Well, as long as I'm not missing some simple tweak to make MySQL run quicker I guess I'll plan to do that. Gavin -- Gavin McCullagh Senior System Administrator IT Services Griffith College South Circular Road Dublin 8 Ireland Tel: +353 1 4163365 http://www.gcd.ie http://www.gcd.ie/brochure.pdf http://www.gcd.ie/opendays This E-mail is from Griffith College. The E-mail and any files transmitted with it are confidential and may be privileged and are intended solely for the use of the individual or entity to whom they are addressed. If you are not the addressee you are prohibited from disclosing its content, copying it or distributing it otherwise than to the addressee. If you have received this e-mail in error, please immediately notify the sender by replying to this e-mail and delete the e-mail from your computer. Bellerophon Ltd, trades as Griffith College (registered in Ireland No. 60469) with its registered address as Griffith College Campus, South Circular Road, Dublin 8, Ireland. -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
Henrik Johansen wrote: I have had about as much of this as I can take now so please, stop spreading FUD about MySQL. Have you used Mysql with datasets in excess of 100-200 million objects? I have. Our current database holds about 400 million File table entries. MySQL requires significant tuning and kernel tweakery, plus uses a lot more memory than postgres does for the same dataset. For Bacula users, it's a lot _easier_ to use Postgres on a large installation than it is to use MySQL. I held off switching to Postgres for a long time because I was unfamiliar with it, however having done so I'm glad that I did - it's required virtually zero tweaking since it was set up and runs approximately twice as fast as MySQL did, with a ram footprint about half the size of MySQL's. Small datasets are fine with MySQL and will probably work better. Ours was brilliant up to about 50 million entries and then required tuning. This discussion is about appropriate tools for the job. If you wish to usefully contribute to the thread then provide some assistance to the OP regarding tuning his MySQL for optimum performance. -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
'Alan Brown' wrote: Henrik Johansen wrote: I have had about as much of this as I can take now so please, stop spreading FUD about MySQL. Have you used Mysql with datasets in excess of 100-200 million objects? Sure - our current Bacula deployment consists of 3 catalog servers with the smallest DB having ~380 million rows. We have other MySQL DB's in production that are considerably larger and so do Facebook, Twitter, Flickr, YouTube, Wikipedia and so on ... I have. Our current database holds about 400 million File table entries. MySQL requires significant tuning and kernel tweakery, plus uses a lot more memory than postgres does for the same dataset. Almost all large MySQL servers we have run Solaris - absolutely no kernel tweaking required. For Bacula users, it's a lot _easier_ to use Postgres on a large installation than it is to use MySQL. Large installations usually have DBA's ? Personally I find it a *lot* easier to apply a few configuration tweaks to a product that I have 8+ years of production experience with than throwing in the towel and starting from scratch with an entirely different product ... I held off switching to Postgres for a long time because I was unfamiliar with it, however having done so I'm glad that I did - it's required virtually zero tweaking since it was set up and runs approximately twice as fast as MySQL did, with a ram footprint about half the size of MySQL's. MySQL, or more specifically InnoDB, needs a bit of love before performing well, I'll admit to that. The upcoming MySQL 5.5 will change much of this however. Small datasets are fine with MySQL and will probably work better. Ours was brilliant up to about 50 million entries and then required tuning. This discussion is about appropriate tools for the job. Yes - and I still consider MySQL to be a highly appropriate tool for the job. Perhaps the MySQL force is particularly strong in me, who knows. If you wish to usefully contribute to the thread then provide some assistance to the OP regarding tuning his MySQL for optimum performance. Re-read the thread - I believe that I already have done so. -- Med venlig hilsen / Best Regards Henrik Johansen hen...@scannet.dk Tlf. 75 53 35 00 ScanNet Group A/S ScanNet -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
On Mon, 08 Nov 2010, Gavin McCullagh wrote: We seem to have the correct indexes on the file table. I've run optimize table and it still takes 14 minutes to build the tree on one of our bigger clients. We have 51 million entries in the file table. I thought I should give some mroe concrete information: I don't suppose this is news to anyone but here's the mysql slow query log to correspond: # Time: 10 14:24:49 # u...@host: bacula[bacula] @ localhost [] # Query_time: 1139.657646 Lock_time: 0.000471 Rows_sent: 4263403 Rows_examined: 50351037 SET timestamp=1289485489; SELECT Path.Path, Filename.Name, Temp.FileIndex, Temp.JobId, LStat, MD5 FROM ( SELECT FileId, Job.JobId AS JobId, FileIndex, File.PathId AS PathId, File.FilenameId AS FilenameId, LStat, MD5 FROM Job, File, ( SELECT MAX(JobTDate) AS JobTDate, PathId, FilenameId FROM ( SELECT JobTDate, PathId, FilenameId FROM File JOIN Job USING (JobId) WHERE File.JobId IN (9944,9950,9973,9996) UNION ALL SELECT JobTDate, PathId, FilenameId FROM BaseFiles JOIN File USING (FileId) JOIN Job ON(BaseJobId = Job.JobId) WHERE BaseFiles.JobId IN (9944,9950,9973,9996) ) AS tmp GROUP BY PathId, FilenameId ) AS T1 WHERE (Job.JobId IN ( SELECT DISTINCT BaseJobId FROM BaseFiles WHERE JobId IN (9944,9950,9973,9996)) OR Job.JobId IN (9944,9950,9973,9996)) AND T1.JobTDate = Job.JobTDate AND Job.JobId = File.JobId AND T1.PathId = File.PathId AND T1.FilenameId = File.FilenameId ) AS Temp JOIN Filename ON (Filename.FilenameId = Temp.FilenameId) JOIN Path ON (Path.PathId = Temp.PathId) WHERE FileIndex 0 ORDER BY Temp.JobId, FileIndex ASC; I've spent some time with the mysqltuner.pl script but to no avail thus far. There's 6GB RAM so it suggests a key buffer size of 4GB which I've set at 4.1GB. This is an Ubuntu Linux server running MySQL v5.1.41. The mysql data is on an MD software RAID 1 array on 7200rpm SATA disks. The tables are MyISAM (which I had understood to be quicker than innodb in low concurrency situations?). The tuner script is suggesting I should disable innodb as we're not using it which I will do though I wouldn't guess that will make a massive difference. There are no fragmented tables currently. Gavin -- Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
On 08/11/10 22:21, Gavin McCullagh wrote: Right you are http://wiki.bacula.org/doku.php?id=faq#restore_takes_a_long_time_to_retrieve_sql_results_from_mysql_catalog There is still an element of move to postgresql though With good reason. I did resist moving to pgsql for quite a while but it does work better. We seem to have the correct indexes on the file table. I've run optimize table and it still takes 14 minutes to build the tree on one of our bigger clients. We have 51 million entries in the file table. Add individual indexes for Fileid, Jobid and Pathid Postgres will work with the combined index for individual table queries, but mysql won't. AB -- The Next 800 Companies to Lead America's Growth: New Video Whitepaper David G. Thomson, author of the best-selling book Blueprint to a Billion shares his insights and actions to help propel your business during the next growth cycle. Listen Now! http://p.sf.net/sfu/SAP-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
On Tue, 09 Nov 2010, Alan Brown wrote: and it still takes 14 minutes to build the tree on one of our bigger clients. We have 51 million entries in the file table. Add individual indexes for Fileid, Jobid and Pathid Postgres will work with the combined index for individual table queries, but mysql won't. The following are the indexes on the file table: mysql SHOW INDEXES FROM File; +---++--+--+-+---+-+--++--++-+ | Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | +---++--+--+-+---+-+--++--++-+ | File | 0 | PRIMARY |1 | FileId | A | 55861148 | NULL | NULL | | BTREE | | | File | 1 | PathId |1 | PathId | A | 735015 | NULL | NULL | | BTREE | | | File | 1 | FilenameId |1 | FilenameId | A | 2539143 | NULL | NULL | | BTREE | | | File | 1 | FilenameId |2 | PathId | A | 13965287 | NULL | NULL | | BTREE | | | File | 1 | JobId|1 | JobId | A | 1324 | NULL | NULL | | BTREE | | | File | 1 | JobId|2 | PathId | A | 2940060 | NULL | NULL | | BTREE | | | File | 1 | JobId|3 | FilenameId | A | 55861148 | NULL | NULL | | BTREE | | | File | 1 | jobid_index |1 | JobId | A | 1324 | NULL | NULL | | BTREE | | | File | 1 | pathid_index |1 | PathId | A | 735015 | NULL | NULL | | BTREE | | +---++--+--+-+---+-+--++--++-+ I added the last two per your instructions. Building the tree took about 14 minutes without these indexes and takes about 17-18 minutes having added them. Have I done something wrong? As FileId is a primary key, it doesn't seem like I should need an extra index on that one -- is that wrong? Thanks Gavin -- The Next 800 Companies to Lead America's Growth: New Video Whitepaper David G. Thomson, author of the best-selling book Blueprint to a Billion shares his insights and actions to help propel your business during the next growth cycle. Listen Now! http://p.sf.net/sfu/SAP-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
Ondrej PLANKA (Ignum profile) wrote: We have several 10+ million file jobs - all run without problem (backup and restore). I am aware of the fact that a lot of Bacula users run PG ( Bacula Systems also does recommend PG for larger setups ) but nevertheless MySQL has served us very well so far. Mysql works well - if tuned, but tuning is a major undertaking when things get large/busy and may take several iterations. The main advantage I've found to Postgress is that when you have a very large database (100+ million entries) it has a far smaller memory/CPU footprint (Half the memory and about 1/4 the cpu load). It's also largely self-tuning - which lets me get on with the business of running backups instead of monitoring database performance... -- The Next 800 Companies to Lead America's Growth: New Video Whitepaper David G. Thomson, author of the best-selling book Blueprint to a Billion shares his insights and actions to help propel your business during the next growth cycle. Listen Now! http://p.sf.net/sfu/SAP-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
On Mon, 08 Nov 2010, Alan Brown wrote: Mysql works well - if tuned, but tuning is a major undertaking when things get large/busy and may take several iterations. Some time back there was an issue with Bacula (v5?) which seemed to come down to a particular query associated (I think) with restores taking a very long time with large datasets on MySQL, but taking a reasonable time on Postgresql. From what I read of the conversation, there didn't seem to be any tuning solution on MySQL. The answer from several people was switch to Postgres. We're using MySQL for Bacula right now but for this reason I've had it in mind to move. Is this still the case or is there a solution now to those MySQL issues. When we do restores, building the tree takes a considerable time now. I haven't had a lot of time to look at it, but suspected it might be down to this issue. Gavin -- The Next 800 Companies to Lead America's Growth: New Video Whitepaper David G. Thomson, author of the best-selling book Blueprint to a Billion shares his insights and actions to help propel your business during the next growth cycle. Listen Now! http://p.sf.net/sfu/SAP-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
Gavin McCullagh wrote: On Mon, 08 Nov 2010, Alan Brown wrote: Mysql works well - if tuned, but tuning is a major undertaking when things get large/busy and may take several iterations. When we do restores, building the tree takes a considerable time now. I haven't had a lot of time to look at it, but suspected it might be down to this issue. That's a classic symptom of not having the right indexes on the File table. This is in the FAQ somewhere and it's the issue you mentioned as affecting mysql but not postgresql (Their index handling is slightly different). The other big contributor to long tree builds is insufficient ram. Keep an eye on swap and the bacula-dir process whilst building the tree. If it starts thrashing, you need more memory. -- The Next 800 Companies to Lead America's Growth: New Video Whitepaper David G. Thomson, author of the best-selling book Blueprint to a Billion shares his insights and actions to help propel your business during the next growth cycle. Listen Now! http://p.sf.net/sfu/SAP-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
Hi Alan, On Mon, 08 Nov 2010, Alan Brown wrote: When we do restores, building the tree takes a considerable time now. I haven't had a lot of time to look at it, but suspected it might be down to this issue. That's a classic symptom of not having the right indexes on the File table. This is in the FAQ somewhere and it's the issue you mentioned as affecting mysql but not postgresql (Their index handling is slightly different). Right you are http://wiki.bacula.org/doku.php?id=faq#restore_takes_a_long_time_to_retrieve_sql_results_from_mysql_catalog There is still an element of move to postgresql though Moving from MySQL to PostgreSQL should make it work much better due to different (more optimized) queries and different SQL engine. http://bugs.bacula.org/view.php?id=1472http://bugs.bacula.org/view.php?id=1472 We seem to have the correct indexes on the file table. I've run optimize table and it still takes 14 minutes to build the tree on one of our bigger clients. We have 51 million entries in the file table. The other big contributor to long tree builds is insufficient ram. Keep an eye on swap and the bacula-dir process whilst building the tree. If it starts thrashing, you need more memory. I think we should be okay on that score but it's something to watch out for. Thanks, Gavin -- The Next 800 Companies to Lead America's Growth: New Video Whitepaper David G. Thomson, author of the best-selling book Blueprint to a Billion shares his insights and actions to help propel your business during the next growth cycle. Listen Now! http://p.sf.net/sfu/SAP-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
'Ondrej PLANKA (Ignum profile)' wrote: Thanks :) Which type of MySQL storage engine are you using? MyISAM or InnoDB for large Bacula system? Can you please copy/paste your MySQL configuration? I mean my.cnf file Please re-read this thread and you should find what you are looking for. Thanks, Ondrej. Henrik Johansen napsal(a): 'Ondrej PLANKA (Ignum profile)' wrote: Hello Henrik, what are you using? MySQL? Yes - all our catalog servers run MySQL. I forgot to mention this in my last post - we are Bacula System customers and they have proved to very supportive and competent. If you are thinking about doing large scale backups with Bacula I can only encourage you to get a support subscription - it is worth every penny. Thanks, Ondrej. 'Mingus Dew' wrote: Henrik, Have you had any problems with slow queries during backup or restore jobs? I'm thinking about http://bugs.bacula.org/view.php?id=1472 specifically, and considering that the bacula.File table already has 73 million rows in it and I haven't even successfully ran the big job yet. Not really. We have several 10+ million file jobs - all run without problem (backup and restore). I am aware of the fact that a lot of Bacula users run PG ( Bacula Systems also does recommend PG for larger setups ) but nevertheless MySQL has served us very well so far. Just curious as a fellow Solaris deployer... Thanks, Shon On Fri, Oct 8, 2010 at 3:30 PM, Henrik Johansen hen...@scannet.dkmailto:hen...@scannet.dk mailto:hen...@scannet.dk%3cmailto:hen...@scannet.dk wrote: 'Mingus Dew' wrote: All, I am running Bacula 5.0.1 on Solaris 10 x86. I'm currently running MySQL 4.1.22 for the database server. I do plan on upgrading to a compatible version of MySQL 5, but migrating to PostgreSQL isn't an option at this time. I am trying to backup to tape a very large number of files for a client. While the data size is manageable at around 2TB, the number of files is incredibly large. The first of the jobs had 27 million files and initially failed because the batch table became Full. I changed the myisam_data_pointer size to a value of 6 in the config. This job was then able to run successfully and did not take too long. I have another job which has 42 million files. I'm not sure what that equates to in rows that need to be inserted, but I can say that I've not been able to successfully run the job, as it seems to hang for over 30 hours in a Dir inserting attributes status. This causes other jobs to backup in the queue and once canceled I have to restart Bacula. I'm looking for way to boost performance of MySQL or Bacula (or both) to get this job completed. You *really* need to upgrade to MySQL 5 and change to InnoDB - there is no way in hell that MySQL 4 + MyISAM is going to perform decent in your situation. Solaris 10 is a Tier 1 platform for MySQL so the latest versions are always available from http://www.mysql.com in the native pkg format so there really is no excuse. We run our Bacula Catalog MySQl servers on Solaris (OpenSolaris) so perhaps I can give you some pointers. Our smallest Bacula DB is currently ~70 GB (381,230,610 rows). Since you are using Solaris 10 I assume that you are going to run MySQL off ZFS - in that case you need to adjust the ZFS recordsize for the filesystem that is going to hold your InnoDB datafiles to match the InnoDB block size. If you are using ZFS you should also consider getting yourself a fast SSD as a SLOG (or to disable the ZIL entirely if you dare) - all InnoDB writes to datafiles are O_SYNC and benefit *greatly* from an SSD in terms of write / transaction speed. If you have enough CPU power to spare you should try turning on compression for the ZFS filesystem holding the datafiles - it also can accelerate DB writes / reads but YMMV. Lastly, our InnoDB related configuration from my.cnf : # InnoDB options skip-innodb_doublewrite innodb_data_home_dir = /tank/db/ innodb_log_group_home_dir = /tank/logs/ innodb_support_xa = false innodb_file_per_table = true innodb_buffer_pool_size = 20G innodb_flush_log_at_trx_commit = 2 innodb_log_buffer_size = 128M innodb_log_file_size = 512M innodb_log_files_in_group = 2 innodb_max_dirty_pages_pct = 90 Thanks, Shon -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.netmailto:Bacula-users@lists.sourceforge.net mailto:bacula-us...@lists.sourceforge.net%3cmailto:Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Med
Re: [Bacula-users] Tuning for large (millions of files) backups?
Am Mon, 01 Nov 2010 06:15:18 +0100 schrieb Ondrej PLANKA (Ignum profile): Thanks :) Which type of MySQL storage engine are you using? MyISAM or InnoDB for large Bacula system? Can you please copy/paste your MySQL configuration? I mean my.cnf file Thanks, Ondrej. I would use InnoDB. a good startingpoint to optimize mysql is http://mysqltuner.pl - Thomas -- Nokia and ATT present the 2010 Calling All Innovators-North America contest Create new apps games for the Nokia N8 for consumers in U.S. and Canada $10 million total in prizes - $4M cash, 500 devices, nearly $6M in marketing Develop with Nokia Qt SDK, Web Runtime, or Java and Publish to Ovi Store http://p.sf.net/sfu/nokia-dev2dev ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
Hello Henrik, what are you using? MySQL? Thanks, Ondrej. 'Mingus Dew' wrote: Henrik, Have you had any problems with slow queries during backup or restore jobs? I'm thinking about http://bugs.bacula.org/view.php?id=1472 specifically, and considering that the bacula.File table already has 73 million rows in it and I haven't even successfully ran the big job yet. Not really. We have several 10+ million file jobs - all run without problem (backup and restore). I am aware of the fact that a lot of Bacula users run PG ( Bacula Systems also does recommend PG for larger setups ) but nevertheless MySQL has served us very well so far. Just curious as a fellow Solaris deployer... Thanks, Shon On Fri, Oct 8, 2010 at 3:30 PM, Henrik Johansen hen...@scannet.dkmailto:hen...@scannet.dk mailto:hen...@scannet.dk%3cmailto:hen...@scannet.dk wrote: 'Mingus Dew' wrote: All, I am running Bacula 5.0.1 on Solaris 10 x86. I'm currently running MySQL 4.1.22 for the database server. I do plan on upgrading to a compatible version of MySQL 5, but migrating to PostgreSQL isn't an option at this time. I am trying to backup to tape a very large number of files for a client. While the data size is manageable at around 2TB, the number of files is incredibly large. The first of the jobs had 27 million files and initially failed because the batch table became Full. I changed the myisam_data_pointer size to a value of 6 in the config. This job was then able to run successfully and did not take too long. I have another job which has 42 million files. I'm not sure what that equates to in rows that need to be inserted, but I can say that I've not been able to successfully run the job, as it seems to hang for over 30 hours in a Dir inserting attributes status. This causes other jobs to backup in the queue and once canceled I have to restart Bacula. I'm looking for way to boost performance of MySQL or Bacula (or both) to get this job completed. You *really* need to upgrade to MySQL 5 and change to InnoDB - there is no way in hell that MySQL 4 + MyISAM is going to perform decent in your situation. Solaris 10 is a Tier 1 platform for MySQL so the latest versions are always available from http://www.mysql.com in the native pkg format so there really is no excuse. We run our Bacula Catalog MySQl servers on Solaris (OpenSolaris) so perhaps I can give you some pointers. Our smallest Bacula DB is currently ~70 GB (381,230,610 rows). Since you are using Solaris 10 I assume that you are going to run MySQL off ZFS - in that case you need to adjust the ZFS recordsize for the filesystem that is going to hold your InnoDB datafiles to match the InnoDB block size. If you are using ZFS you should also consider getting yourself a fast SSD as a SLOG (or to disable the ZIL entirely if you dare) - all InnoDB writes to datafiles are O_SYNC and benefit *greatly* from an SSD in terms of write / transaction speed. If you have enough CPU power to spare you should try turning on compression for the ZFS filesystem holding the datafiles - it also can accelerate DB writes / reads but YMMV. Lastly, our InnoDB related configuration from my.cnf : # InnoDB options skip-innodb_doublewrite innodb_data_home_dir = /tank/db/ innodb_log_group_home_dir = /tank/logs/ innodb_support_xa = false innodb_file_per_table = true innodb_buffer_pool_size = 20G innodb_flush_log_at_trx_commit = 2 innodb_log_buffer_size = 128M innodb_log_file_size = 512M innodb_log_files_in_group = 2 innodb_max_dirty_pages_pct = 90 Thanks, Shon -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.netmailto:Bacula-users@lists.sourceforge.net mailto:bacula-us...@lists.sourceforge.net%3cmailto:Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Med venlig hilsen / Best Regards Henrik Johansen hen...@scannet.dkmailto:hen...@scannet.dk mailto:hen...@scannet.dk%3cmailto:hen...@scannet.dk Tlf. 75 53 35 00 ScanNet Group A/S ScanNet -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net mailto:Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Med venlig hilsen / Best
Re: [Bacula-users] Tuning for large (millions of files) backups?
'Ondrej PLANKA (Ignum profile)' wrote: Hello Henrik, what are you using? MySQL? Yes - all our catalog servers run MySQL. I forgot to mention this in my last post - we are Bacula System customers and they have proved to very supportive and competent. If you are thinking about doing large scale backups with Bacula I can only encourage you to get a support subscription - it is worth every penny. Thanks, Ondrej. 'Mingus Dew' wrote: Henrik, Have you had any problems with slow queries during backup or restore jobs? I'm thinking about http://bugs.bacula.org/view.php?id=1472 specifically, and considering that the bacula.File table already has 73 million rows in it and I haven't even successfully ran the big job yet. Not really. We have several 10+ million file jobs - all run without problem (backup and restore). I am aware of the fact that a lot of Bacula users run PG ( Bacula Systems also does recommend PG for larger setups ) but nevertheless MySQL has served us very well so far. Just curious as a fellow Solaris deployer... Thanks, Shon On Fri, Oct 8, 2010 at 3:30 PM, Henrik Johansen hen...@scannet.dkmailto:hen...@scannet.dk mailto:hen...@scannet.dk%3cmailto:hen...@scannet.dk wrote: 'Mingus Dew' wrote: All, I am running Bacula 5.0.1 on Solaris 10 x86. I'm currently running MySQL 4.1.22 for the database server. I do plan on upgrading to a compatible version of MySQL 5, but migrating to PostgreSQL isn't an option at this time. I am trying to backup to tape a very large number of files for a client. While the data size is manageable at around 2TB, the number of files is incredibly large. The first of the jobs had 27 million files and initially failed because the batch table became Full. I changed the myisam_data_pointer size to a value of 6 in the config. This job was then able to run successfully and did not take too long. I have another job which has 42 million files. I'm not sure what that equates to in rows that need to be inserted, but I can say that I've not been able to successfully run the job, as it seems to hang for over 30 hours in a Dir inserting attributes status. This causes other jobs to backup in the queue and once canceled I have to restart Bacula. I'm looking for way to boost performance of MySQL or Bacula (or both) to get this job completed. You *really* need to upgrade to MySQL 5 and change to InnoDB - there is no way in hell that MySQL 4 + MyISAM is going to perform decent in your situation. Solaris 10 is a Tier 1 platform for MySQL so the latest versions are always available from http://www.mysql.com in the native pkg format so there really is no excuse. We run our Bacula Catalog MySQl servers on Solaris (OpenSolaris) so perhaps I can give you some pointers. Our smallest Bacula DB is currently ~70 GB (381,230,610 rows). Since you are using Solaris 10 I assume that you are going to run MySQL off ZFS - in that case you need to adjust the ZFS recordsize for the filesystem that is going to hold your InnoDB datafiles to match the InnoDB block size. If you are using ZFS you should also consider getting yourself a fast SSD as a SLOG (or to disable the ZIL entirely if you dare) - all InnoDB writes to datafiles are O_SYNC and benefit *greatly* from an SSD in terms of write / transaction speed. If you have enough CPU power to spare you should try turning on compression for the ZFS filesystem holding the datafiles - it also can accelerate DB writes / reads but YMMV. Lastly, our InnoDB related configuration from my.cnf : # InnoDB options skip-innodb_doublewrite innodb_data_home_dir = /tank/db/ innodb_log_group_home_dir = /tank/logs/ innodb_support_xa = false innodb_file_per_table = true innodb_buffer_pool_size = 20G innodb_flush_log_at_trx_commit = 2 innodb_log_buffer_size = 128M innodb_log_file_size = 512M innodb_log_files_in_group = 2 innodb_max_dirty_pages_pct = 90 Thanks, Shon -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.netmailto:Bacula-users@lists.sourceforge.net mailto:bacula-us...@lists.sourceforge.net%3cmailto:Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Med venlig hilsen / Best Regards Henrik Johansen hen...@scannet.dkmailto:hen...@scannet.dk mailto:hen...@scannet.dk%3cmailto:hen...@scannet.dk Tlf. 75 53 35 00 ScanNet Group A/S ScanNet -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code
Re: [Bacula-users] Tuning for large (millions of files) backups?
Thanks :) Which type of MySQL storage engine are you using? MyISAM or InnoDB for large Bacula system? Can you please copy/paste your MySQL configuration? I mean my.cnf file Thanks, Ondrej. Henrik Johansen napsal(a): 'Ondrej PLANKA (Ignum profile)' wrote: Hello Henrik, what are you using? MySQL? Yes - all our catalog servers run MySQL. I forgot to mention this in my last post - we are Bacula System customers and they have proved to very supportive and competent. If you are thinking about doing large scale backups with Bacula I can only encourage you to get a support subscription - it is worth every penny. Thanks, Ondrej. 'Mingus Dew' wrote: Henrik, Have you had any problems with slow queries during backup or restore jobs? I'm thinking about http://bugs.bacula.org/view.php?id=1472 specifically, and considering that the bacula.File table already has 73 million rows in it and I haven't even successfully ran the big job yet. Not really. We have several 10+ million file jobs - all run without problem (backup and restore). I am aware of the fact that a lot of Bacula users run PG ( Bacula Systems also does recommend PG for larger setups ) but nevertheless MySQL has served us very well so far. Just curious as a fellow Solaris deployer... Thanks, Shon On Fri, Oct 8, 2010 at 3:30 PM, Henrik Johansen hen...@scannet.dkmailto:hen...@scannet.dk mailto:hen...@scannet.dk%3cmailto:hen...@scannet.dk wrote: 'Mingus Dew' wrote: All, I am running Bacula 5.0.1 on Solaris 10 x86. I'm currently running MySQL 4.1.22 for the database server. I do plan on upgrading to a compatible version of MySQL 5, but migrating to PostgreSQL isn't an option at this time. I am trying to backup to tape a very large number of files for a client. While the data size is manageable at around 2TB, the number of files is incredibly large. The first of the jobs had 27 million files and initially failed because the batch table became Full. I changed the myisam_data_pointer size to a value of 6 in the config. This job was then able to run successfully and did not take too long. I have another job which has 42 million files. I'm not sure what that equates to in rows that need to be inserted, but I can say that I've not been able to successfully run the job, as it seems to hang for over 30 hours in a Dir inserting attributes status. This causes other jobs to backup in the queue and once canceled I have to restart Bacula. I'm looking for way to boost performance of MySQL or Bacula (or both) to get this job completed. You *really* need to upgrade to MySQL 5 and change to InnoDB - there is no way in hell that MySQL 4 + MyISAM is going to perform decent in your situation. Solaris 10 is a Tier 1 platform for MySQL so the latest versions are always available from http://www.mysql.com in the native pkg format so there really is no excuse. We run our Bacula Catalog MySQl servers on Solaris (OpenSolaris) so perhaps I can give you some pointers. Our smallest Bacula DB is currently ~70 GB (381,230,610 rows). Since you are using Solaris 10 I assume that you are going to run MySQL off ZFS - in that case you need to adjust the ZFS recordsize for the filesystem that is going to hold your InnoDB datafiles to match the InnoDB block size. If you are using ZFS you should also consider getting yourself a fast SSD as a SLOG (or to disable the ZIL entirely if you dare) - all InnoDB writes to datafiles are O_SYNC and benefit *greatly* from an SSD in terms of write / transaction speed. If you have enough CPU power to spare you should try turning on compression for the ZFS filesystem holding the datafiles - it also can accelerate DB writes / reads but YMMV. Lastly, our InnoDB related configuration from my.cnf : # InnoDB options skip-innodb_doublewrite innodb_data_home_dir = /tank/db/ innodb_log_group_home_dir = /tank/logs/ innodb_support_xa = false innodb_file_per_table = true innodb_buffer_pool_size = 20G innodb_flush_log_at_trx_commit = 2 innodb_log_buffer_size = 128M innodb_log_file_size = 512M innodb_log_files_in_group = 2 innodb_max_dirty_pages_pct = 90 Thanks, Shon -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.netmailto:Bacula-users@lists.sourceforge.net mailto:bacula-us...@lists.sourceforge.net%3cmailto:Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Med venlig hilsen / Best Regards Henrik Johansen
Re: [Bacula-users] Tuning for large (millions of files) backups?
Henrik, Have you had any problems with slow queries during backup or restore jobs? I'm thinking about http://bugs.bacula.org/view.php?id=1472specifically, and considering that the bacula.File table already has 73 million rows in it and I haven't even successfully ran the big job yet. Just curious as a fellow Solaris deployer... Thanks, Shon On Fri, Oct 8, 2010 at 3:30 PM, Henrik Johansen hen...@scannet.dk wrote: 'Mingus Dew' wrote: All, I am running Bacula 5.0.1 on Solaris 10 x86. I'm currently running MySQL 4.1.22 for the database server. I do plan on upgrading to a compatible version of MySQL 5, but migrating to PostgreSQL isn't an option at this time. I am trying to backup to tape a very large number of files for a client. While the data size is manageable at around 2TB, the number of files is incredibly large. The first of the jobs had 27 million files and initially failed because the batch table became Full. I changed the myisam_data_pointer size to a value of 6 in the config. This job was then able to run successfully and did not take too long. I have another job which has 42 million files. I'm not sure what that equates to in rows that need to be inserted, but I can say that I've not been able to successfully run the job, as it seems to hang for over 30 hours in a Dir inserting attributes status. This causes other jobs to backup in the queue and once canceled I have to restart Bacula. I'm looking for way to boost performance of MySQL or Bacula (or both) to get this job completed. You *really* need to upgrade to MySQL 5 and change to InnoDB - there is no way in hell that MySQL 4 + MyISAM is going to perform decent in your situation. Solaris 10 is a Tier 1 platform for MySQL so the latest versions are always available from www.mysql.com in the native pkg format so there really is no excuse. We run our Bacula Catalog MySQl servers on Solaris (OpenSolaris) so perhaps I can give you some pointers. Our smallest Bacula DB is currently ~70 GB (381,230,610 rows). Since you are using Solaris 10 I assume that you are going to run MySQL off ZFS - in that case you need to adjust the ZFS recordsize for the filesystem that is going to hold your InnoDB datafiles to match the InnoDB block size. If you are using ZFS you should also consider getting yourself a fast SSD as a SLOG (or to disable the ZIL entirely if you dare) - all InnoDB writes to datafiles are O_SYNC and benefit *greatly* from an SSD in terms of write / transaction speed. If you have enough CPU power to spare you should try turning on compression for the ZFS filesystem holding the datafiles - it also can accelerate DB writes / reads but YMMV. Lastly, our InnoDB related configuration from my.cnf : # InnoDB options skip-innodb_doublewrite innodb_data_home_dir = /tank/db/ innodb_log_group_home_dir = /tank/logs/ innodb_support_xa = false innodb_file_per_table = true innodb_buffer_pool_size = 20G innodb_flush_log_at_trx_commit = 2 innodb_log_buffer_size = 128M innodb_log_file_size = 512M innodb_log_files_in_group = 2 innodb_max_dirty_pages_pct = 90 Thanks, Shon -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Med venlig hilsen / Best Regards Henrik Johansen hen...@scannet.dk Tlf. 75 53 35 00 ScanNet Group A/S ScanNet -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
'Mingus Dew' wrote: Henrik, Have you had any problems with slow queries during backup or restore jobs? I'm thinking about http://bugs.bacula.org/view.php?id=1472 specifically, and considering that the bacula.File table already has 73 million rows in it and I haven't even successfully ran the big job yet. Not really. We have several 10+ million file jobs - all run without problem (backup and restore). I am aware of the fact that a lot of Bacula users run PG ( Bacula Systems also does recommend PG for larger setups ) but nevertheless MySQL has served us very well so far. Just curious as a fellow Solaris deployer... Thanks, Shon On Fri, Oct 8, 2010 at 3:30 PM, Henrik Johansen hen...@scannet.dkmailto:hen...@scannet.dk wrote: 'Mingus Dew' wrote: All, I am running Bacula 5.0.1 on Solaris 10 x86. I'm currently running MySQL 4.1.22 for the database server. I do plan on upgrading to a compatible version of MySQL 5, but migrating to PostgreSQL isn't an option at this time. I am trying to backup to tape a very large number of files for a client. While the data size is manageable at around 2TB, the number of files is incredibly large. The first of the jobs had 27 million files and initially failed because the batch table became Full. I changed the myisam_data_pointer size to a value of 6 in the config. This job was then able to run successfully and did not take too long. I have another job which has 42 million files. I'm not sure what that equates to in rows that need to be inserted, but I can say that I've not been able to successfully run the job, as it seems to hang for over 30 hours in a Dir inserting attributes status. This causes other jobs to backup in the queue and once canceled I have to restart Bacula. I'm looking for way to boost performance of MySQL or Bacula (or both) to get this job completed. You *really* need to upgrade to MySQL 5 and change to InnoDB - there is no way in hell that MySQL 4 + MyISAM is going to perform decent in your situation. Solaris 10 is a Tier 1 platform for MySQL so the latest versions are always available from www.mysql.com in the native pkg format so there really is no excuse. We run our Bacula Catalog MySQl servers on Solaris (OpenSolaris) so perhaps I can give you some pointers. Our smallest Bacula DB is currently ~70 GB (381,230,610 rows). Since you are using Solaris 10 I assume that you are going to run MySQL off ZFS - in that case you need to adjust the ZFS recordsize for the filesystem that is going to hold your InnoDB datafiles to match the InnoDB block size. If you are using ZFS you should also consider getting yourself a fast SSD as a SLOG (or to disable the ZIL entirely if you dare) - all InnoDB writes to datafiles are O_SYNC and benefit *greatly* from an SSD in terms of write / transaction speed. If you have enough CPU power to spare you should try turning on compression for the ZFS filesystem holding the datafiles - it also can accelerate DB writes / reads but YMMV. Lastly, our InnoDB related configuration from my.cnf : # InnoDB options skip-innodb_doublewrite innodb_data_home_dir = /tank/db/ innodb_log_group_home_dir = /tank/logs/ innodb_support_xa = false innodb_file_per_table = true innodb_buffer_pool_size = 20G innodb_flush_log_at_trx_commit = 2 innodb_log_buffer_size = 128M innodb_log_file_size = 512M innodb_log_files_in_group = 2 innodb_max_dirty_pages_pct = 90 Thanks, Shon -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.netmailto:Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Med venlig hilsen / Best Regards Henrik Johansen hen...@scannet.dkmailto:hen...@scannet.dk Tlf. 75 53 35 00 ScanNet Group A/S ScanNet -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Med venlig hilsen / Best Regards Henrik Johansen hen...@scannet.dk Tlf. 75 53 35 00 ScanNet Group A/S ScanNet -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend
Re: [Bacula-users] Tuning for large (millions of files) backups?
Alan Brown wrote: You are going to hit a big pain point with myisam with that many files anyway (it breaks around 4 billion entries without tuning), but even inno will grow large/slow and need a lot of my.cnf tuning That should be 4Gb - about 50 million entries. -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
Bruno Friedmann wrote: Rude answer : If you really want to use Mysql drop the myisam to innodb. But you don't want to use mysql for that job, just use Postgresql fine tuned with batch insert enabled. Seconded - having been through this issue. You are going to hit a big pain point with myisam with that many files anyway (it breaks around 4 billion entries without tuning), but even inno will grow large/slow and need a lot of my.cnf tuning Go straight to Postgres - you'll need it eventually anyway, then read up on tuning it. For large databases it runs faster and uses less memory. -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
On 12/10/10, Alan Brown (a...@mssl.ucl.ac.uk) wrote: Bruno Friedmann wrote: But you don't want to use mysql for that job, just use Postgresql fine tuned with batch insert enabled. Seconded - having been through this issue. I am running Postgresql with batch insert with jobs of around 8 million files, and it works without any problems. Postgresql is tremendous at providing a smooth upgrade path too. We migrated a lot of services with few problems from each major release starting in the low 7.x release series. Regards Rory -- Rory Campbell-Lange r...@campbell-lange.net Campbell-Lange Workshop www.campbell-lange.net 0207 6311 555 3 Tottenham Street London W1T 2AF Registered in England No. 04551928 -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
Henrik, I really appreciate your reply, particularly as a fellow Bacula-on-Solaris user. I do not have my databases on ZFS, only my Bacula storage. I'll probably have to tune for local disk. Thanks very much, Shon On Fri, Oct 8, 2010 at 3:30 PM, Henrik Johansen hen...@scannet.dk wrote: 'Mingus Dew' wrote: All, I am running Bacula 5.0.1 on Solaris 10 x86. I'm currently running MySQL 4.1.22 for the database server. I do plan on upgrading to a compatible version of MySQL 5, but migrating to PostgreSQL isn't an option at this time. I am trying to backup to tape a very large number of files for a client. While the data size is manageable at around 2TB, the number of files is incredibly large. The first of the jobs had 27 million files and initially failed because the batch table became Full. I changed the myisam_data_pointer size to a value of 6 in the config. This job was then able to run successfully and did not take too long. I have another job which has 42 million files. I'm not sure what that equates to in rows that need to be inserted, but I can say that I've not been able to successfully run the job, as it seems to hang for over 30 hours in a Dir inserting attributes status. This causes other jobs to backup in the queue and once canceled I have to restart Bacula. I'm looking for way to boost performance of MySQL or Bacula (or both) to get this job completed. You *really* need to upgrade to MySQL 5 and change to InnoDB - there is no way in hell that MySQL 4 + MyISAM is going to perform decent in your situation. Solaris 10 is a Tier 1 platform for MySQL so the latest versions are always available from www.mysql.com in the native pkg format so there really is no excuse. We run our Bacula Catalog MySQl servers on Solaris (OpenSolaris) so perhaps I can give you some pointers. Our smallest Bacula DB is currently ~70 GB (381,230,610 rows). Since you are using Solaris 10 I assume that you are going to run MySQL off ZFS - in that case you need to adjust the ZFS recordsize for the filesystem that is going to hold your InnoDB datafiles to match the InnoDB block size. If you are using ZFS you should also consider getting yourself a fast SSD as a SLOG (or to disable the ZIL entirely if you dare) - all InnoDB writes to datafiles are O_SYNC and benefit *greatly* from an SSD in terms of write / transaction speed. If you have enough CPU power to spare you should try turning on compression for the ZFS filesystem holding the datafiles - it also can accelerate DB writes / reads but YMMV. Lastly, our InnoDB related configuration from my.cnf : # InnoDB options skip-innodb_doublewrite innodb_data_home_dir = /tank/db/ innodb_log_group_home_dir = /tank/logs/ innodb_support_xa = false innodb_file_per_table = true innodb_buffer_pool_size = 20G innodb_flush_log_at_trx_commit = 2 innodb_log_buffer_size = 128M innodb_log_file_size = 512M innodb_log_files_in_group = 2 innodb_max_dirty_pages_pct = 90 Thanks, Shon -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Med venlig hilsen / Best Regards Henrik Johansen hen...@scannet.dk Tlf. 75 53 35 00 ScanNet Group A/S ScanNet -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
On 10/07/2010 11:03 PM, Mingus Dew wrote: All, I am running Bacula 5.0.1 on Solaris 10 x86. I'm currently running MySQL 4.1.22 for the database server. I do plan on upgrading to a compatible version of MySQL 5, but migrating to PostgreSQL isn't an option at this time. I am trying to backup to tape a very large number of files for a client. While the data size is manageable at around 2TB, the number of files is incredibly large. The first of the jobs had 27 million files and initially failed because the batch table became Full. I changed the myisam_data_pointer size to a value of 6 in the config. This job was then able to run successfully and did not take too long. I have another job which has 42 million files. I'm not sure what that equates to in rows that need to be inserted, but I can say that I've not been able to successfully run the job, as it seems to hang for over 30 hours in a Dir inserting attributes status. This causes other jobs to backup in the queue and once canceled I have to restart Bacula. I'm looking for way to boost performance of MySQL or Bacula (or both) to get this job completed. Thanks, Shon Rude answer : If you really want to use Mysql drop the myisam to innodb. But you don't want to use mysql for that job, just use Postgresql fine tuned with batch insert enabled. :-) -- Bruno Friedmann (irc:tigerfoot) Ioda-Net Sàrl www.ioda-net.ch openSUSE Member User www.ioda.net/r/osu Blog www.ioda.net/r/blog fsfe fellowship www.fsfe.org GPG KEY : D5C9B751C4653227 vcard : http://it.ioda-net.ch/ioda-net.vcf -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
Bruno, Not so rude at all :) You've made me think of 2 questions How difficult is it (or procedure for) converting to InnoDB and what exactly will this gain in performance increase? Also, you mention Postgresql and batch inserts. Does Bacula not use batch inserts with MySQL by default? I'm assuming I'm using batch inserts because Bacula uses a table called 'batch' -Shon On Fri, Oct 8, 2010 at 2:07 AM, Bruno Friedmann br...@ioda-net.ch wrote: On 10/07/2010 11:03 PM, Mingus Dew wrote: All, I am running Bacula 5.0.1 on Solaris 10 x86. I'm currently running MySQL 4.1.22 for the database server. I do plan on upgrading to a compatible version of MySQL 5, but migrating to PostgreSQL isn't an option at this time. I am trying to backup to tape a very large number of files for a client. While the data size is manageable at around 2TB, the number of files is incredibly large. The first of the jobs had 27 million files and initially failed because the batch table became Full. I changed the myisam_data_pointer size to a value of 6 in the config. This job was then able to run successfully and did not take too long. I have another job which has 42 million files. I'm not sure what that equates to in rows that need to be inserted, but I can say that I've not been able to successfully run the job, as it seems to hang for over 30 hours in a Dir inserting attributes status. This causes other jobs to backup in the queue and once canceled I have to restart Bacula. I'm looking for way to boost performance of MySQL or Bacula (or both) to get this job completed. Thanks, Shon Rude answer : If you really want to use Mysql drop the myisam to innodb. But you don't want to use mysql for that job, just use Postgresql fine tuned with batch insert enabled. :-) -- Bruno Friedmann (irc:tigerfoot) Ioda-Net Sàrl www.ioda-net.ch openSUSE Member User www.ioda.net/r/osu Blog www.ioda.net/r/blog fsfe fellowship www.fsfe.org GPG KEY : D5C9B751C4653227 vcard : http://it.ioda-net.ch/ioda-net.vcf -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
For batch insert by default on mysql, it could be or not, depending on several factors is mysql pthread safe or not, and configure option choose during building time. The mysql 4 is obsolete now with 5.0.3 (I think there's some good reasons for that) Transforming table to innodb is quite simple, but depend how much and how are the indexes on tables. Innodb doesn't like varchar indexes 254. So if you don't have any blockers, you will have to upgrade your mysql to something more recent, so perhaps moving to postgresql is a perfect alternative, just in time. For performances even with batch-inserted, I'm remembering a graph that Eric B have show during prototyping db last year at pgdays.eu ( perharps the slides are always there ) Eric ? any suggestions if you read that. Anyway, I'm not using Solaris, so it would be nice to have some advice from experimented people on that plateform. On 10/08/2010 02:36 PM, Mingus Dew wrote: Bruno, Not so rude at all :) You've made me think of 2 questions How difficult is it (or procedure for) converting to InnoDB and what exactly will this gain in performance increase? Also, you mention Postgresql and batch inserts. Does Bacula not use batch inserts with MySQL by default? I'm assuming I'm using batch inserts because Bacula uses a table called 'batch' -Shon On Fri, Oct 8, 2010 at 2:07 AM, Bruno Friedmann br...@ioda-net.ch wrote: On 10/07/2010 11:03 PM, Mingus Dew wrote: All, I am running Bacula 5.0.1 on Solaris 10 x86. I'm currently running MySQL 4.1.22 for the database server. I do plan on upgrading to a compatible version of MySQL 5, but migrating to PostgreSQL isn't an option at this time. I am trying to backup to tape a very large number of files for a client. While the data size is manageable at around 2TB, the number of files is incredibly large. The first of the jobs had 27 million files and initially failed because the batch table became Full. I changed the myisam_data_pointer size to a value of 6 in the config. This job was then able to run successfully and did not take too long. I have another job which has 42 million files. I'm not sure what that equates to in rows that need to be inserted, but I can say that I've not been able to successfully run the job, as it seems to hang for over 30 hours in a Dir inserting attributes status. This causes other jobs to backup in the queue and once canceled I have to restart Bacula. I'm looking for way to boost performance of MySQL or Bacula (or both) to get this job completed. Thanks, Shon Rude answer : If you really want to use Mysql drop the myisam to innodb. But you don't want to use mysql for that job, just use Postgresql fine tuned with batch insert enabled. :-) -- Bruno Friedmann (irc:tigerfoot) Ioda-Net Sàrl www.ioda-net.ch openSUSE Member User www.ioda.net/r/osu Blog www.ioda.net/r/blog fsfe fellowship www.fsfe.org GPG KEY : D5C9B751C4653227 vcard : http://it.ioda-net.ch/ioda-net.vcf -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Bruno Friedmann (irc:tigerfoot) Ioda-Net Sàrl www.ioda-net.ch openSUSE Member User www.ioda.net/r/osu Blog www.ioda.net/r/blog fsfe fellowship www.fsfe.org GPG KEY : D5C9B751C4653227 vcard : http://it.ioda-net.ch/ioda-net.vcf -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
'Mingus Dew' wrote: All, I am running Bacula 5.0.1 on Solaris 10 x86. I'm currently running MySQL 4.1.22 for the database server. I do plan on upgrading to a compatible version of MySQL 5, but migrating to PostgreSQL isn't an option at this time. I am trying to backup to tape a very large number of files for a client. While the data size is manageable at around 2TB, the number of files is incredibly large. The first of the jobs had 27 million files and initially failed because the batch table became Full. I changed the myisam_data_pointer size to a value of 6 in the config. This job was then able to run successfully and did not take too long. I have another job which has 42 million files. I'm not sure what that equates to in rows that need to be inserted, but I can say that I've not been able to successfully run the job, as it seems to hang for over 30 hours in a Dir inserting attributes status. This causes other jobs to backup in the queue and once canceled I have to restart Bacula. I'm looking for way to boost performance of MySQL or Bacula (or both) to get this job completed. You *really* need to upgrade to MySQL 5 and change to InnoDB - there is no way in hell that MySQL 4 + MyISAM is going to perform decent in your situation. Solaris 10 is a Tier 1 platform for MySQL so the latest versions are always available from www.mysql.com in the native pkg format so there really is no excuse. We run our Bacula Catalog MySQl servers on Solaris (OpenSolaris) so perhaps I can give you some pointers. Our smallest Bacula DB is currently ~70 GB (381,230,610 rows). Since you are using Solaris 10 I assume that you are going to run MySQL off ZFS - in that case you need to adjust the ZFS recordsize for the filesystem that is going to hold your InnoDB datafiles to match the InnoDB block size. If you are using ZFS you should also consider getting yourself a fast SSD as a SLOG (or to disable the ZIL entirely if you dare) - all InnoDB writes to datafiles are O_SYNC and benefit *greatly* from an SSD in terms of write / transaction speed. If you have enough CPU power to spare you should try turning on compression for the ZFS filesystem holding the datafiles - it also can accelerate DB writes / reads but YMMV. Lastly, our InnoDB related configuration from my.cnf : # InnoDB options skip-innodb_doublewrite innodb_data_home_dir = /tank/db/ innodb_log_group_home_dir = /tank/logs/ innodb_support_xa = false innodb_file_per_table = true innodb_buffer_pool_size = 20G innodb_flush_log_at_trx_commit = 2 innodb_log_buffer_size = 128M innodb_log_file_size = 512M innodb_log_files_in_group = 2 innodb_max_dirty_pages_pct = 90 Thanks, Shon -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- Med venlig hilsen / Best Regards Henrik Johansen hen...@scannet.dk Tlf. 75 53 35 00 ScanNet Group A/S ScanNet -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
On 10/08/10 15:30, Henrik Johansen wrote: Since you are using Solaris 10 I assume that you are going to run MySQL off ZFS - in that case you need to adjust the ZFS recordsize for the filesystem that is going to hold your InnoDB datafiles to match the InnoDB block size. Henrik, This is an interesting observation. How does one determine/set the InnoDB block size? -- Phil Stracchino, CDK#2 DoD#299792458 ICBM: 43.5607, -71.355 ala...@caerllewys.net ala...@metrocast.net p...@co.ordinate.org Renaissance Man, Unix ronin, Perl hacker, Free Stater It's not the years, it's the mileage. -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
This is an interesting observation. How does one determine/set the InnoDB block size? Sorry for butting in here, but I've been following this thread. You can't change the InnoDB block size unless you recompile from source, from what I understand...but that's besides the point. Using InnoDB adds quite a bit of overhead to most database operations; shouldn't Bacula be using MyISAM tables, which are much faster? My thinking is that there is not a lot of concurrency with database reads and writes, and probably not much need for referential integrity...or am I missing something? Tim Gustafson Baskin School of Engineering UC Santa Cruz t...@soe.ucsc.edu 831-459-5354 -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
Phil Stracchino wrote: On 10/08/10 15:30, Henrik Johansen wrote: Since you are using Solaris 10 I assume that you are going to run MySQL off ZFS - in that case you need to adjust the ZFS recordsize for the filesystem that is going to hold your InnoDB datafiles to match the InnoDB block size. Henrik, This is an interesting observation. How does one determine/set the InnoDB block size? Phil, please see http://dev.mysql.com/tech-resources/articles/mysql-zfs.html#Set_the_ZFS_Recordsize_to_match_the_block_size 16K is the zfs recodesize I'm using. Attila -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Tuning for large (millions of files) backups?
On 10/08/10 17:49, Attila Fülöp wrote: please see http://dev.mysql.com/tech-resources/articles/mysql-zfs.html#Set_the_ZFS_Recordsize_to_match_the_block_size 16K is the zfs recodesize I'm using. Aha! Thanks, Attila. Exactly what I needed. -- Phil Stracchino, CDK#2 DoD#299792458 ICBM: 43.5607, -71.355 ala...@caerllewys.net ala...@metrocast.net p...@co.ordinate.org Renaissance Man, Unix ronin, Perl hacker, Free Stater It's not the years, it's the mileage. -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
[Bacula-users] Tuning for large (millions of files) backups?
All, I am running Bacula 5.0.1 on Solaris 10 x86. I'm currently running MySQL 4.1.22 for the database server. I do plan on upgrading to a compatible version of MySQL 5, but migrating to PostgreSQL isn't an option at this time. I am trying to backup to tape a very large number of files for a client. While the data size is manageable at around 2TB, the number of files is incredibly large. The first of the jobs had 27 million files and initially failed because the batch table became Full. I changed the myisam_data_pointer size to a value of 6 in the config. This job was then able to run successfully and did not take too long. I have another job which has 42 million files. I'm not sure what that equates to in rows that need to be inserted, but I can say that I've not been able to successfully run the job, as it seems to hang for over 30 hours in a Dir inserting attributes status. This causes other jobs to backup in the queue and once canceled I have to restart Bacula. I'm looking for way to boost performance of MySQL or Bacula (or both) to get this job completed. Thanks, Shon -- Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users