Re: [Bacula-users] EXTERNAL - Re: /var/lib/mysql at 100%
Just wanted to say thank you Phil for the pointers on this. I have moved the database directory for now and working on increasing the database filesystem disk space. Cheers Chaz Chaz Vidal | ICT Infrastructure | Tel: +61-8-8128-4397 | Mob: +61-492-874-982 | chaz.vi...@sahmri.com -Original Message- From: Chaz Vidal Sent: Monday, 6 July 2020 2:22 PM To: Phil Stracchino Cc: bacula-users Subject: Re: [Bacula-users] EXTERNAL - Re: /var/lib/mysql at 100% Very very much appreciated Phil! I'll schedule something to fix this up and we will ensure we remediate this system. I'll let you know how we go. Cheers Chaz -Original Message- From: Phil Stracchino Sent: Monday, 6 July 2020 12:10 PM To: Chaz Vidal Cc: bacula-users Subject: Re: EXTERNAL - Re: [Bacula-users] /var/lib/mysql at 100% On 2020-07-05 19:45, Chaz Vidal wrote: > Thanks Phil, > For some reason my backups are still running and completing. Is it because > we still have table space? Probably, yes. It most likely means you have enough free space in the tablespace to keep going for now. > This is an ext4 filesystem. I have attempted a dump whilst bacula was running > and the resulting dump file was about 53% (70GB) of the existing Bacula db > directory which is 130GB. OK, but that does not mean that you have 60GB of free space to play with. Your dump contains only the data and schemas of your database, not the contents of the table indexes or unused column space. The indexes will be rebuilt by mysqld when you reload the data from your dump. A good set of indexes on a large database consumes significant space. but it is the voluminous and rapidly-traversable indexes that allow you to retrieve data from your database quickly. Indexes are at the heart of what makes a relational database better and faster than a flat file. They are what tells the database engine exactly where in your 70GB of data a single specific piece of data is stored. > > The spool directory is so much larger with 6TB available. Is this a > potential solution? > > I'm starting out with Bacula so I'm assuming that we do a "/etc/init.d/bacula > stop", run the database dump commands and then start? If you're referring to the spool directory for job spooling, it might be unwise to also put your MariaDB data directory there unless you first change your high-water setting to make sure that spooled jobs cannot encroach on your database space. With that proviso, you could do it as a temporary measure until you can make more permanent provisions to expand your database space. Here's the fastest way to accomplish the move: 1. Stop Bacula 2. Stop MariaDB 3. Create your new data directory 4. Move or copy the entire contents of /var/lib/mysql to the new directory. Make sure the new directory has the same ownership and permissions as /var/lib/mysql. 5. Edit the global MariaDB configuration files to point datadir to the new location. Also update anything else in the MariaDB configuration that refers to /var/lib/mysql (tmpdir for instance). 6. Start MariaDB and make sure it comes up correctly 7. Edit your Bacula configuration files to set a safe spool high-water mark 8. Restart Bacula This does not touch on any optimization, tuning, enabling file-per-table, or getting rid of the unwanted lost+found in the data directory, but it will get you back up and running much more quickly than reloading 70GB of data from a logical dump and does not risk missing anything during a dump-and-reload. You can prepare plans to deal with those issues later. Remember that this is a stopgap solution until you can permanently expand your dedicated database storage. A couple of things to remember there: — XFS is preferred over ext* for performance reasons (in fact, further development of the ext filesystem has been officially abandoned by Red Hat); — The MySQL/.MariaDB data directory does not have to be located at /var/lib/mysql, that's just the default; — It is advised NOT to use the root of an ext* volume as your data directory (for example, if your data volume is ext4 mounted at /mariadb, make your data directory something like /mariadb/data) — Whenever possible, database storage should be on physically separate devices from OS loads or Bacula usage to minimize I/O contention on the database. -- Phil Stracchino Babylon Communications ph...@caerllewys.net p...@co.ordinate.org Landline: +1.603.293.8485 Mobile: +1.603.998.6958 ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Community binaries on Fedora 31: libreadline.so.7 not available
On 2020-07-08 02:54, Davide Franco wrote: > Hello Phil, > > You dont need a dev environment as I have uploaded Bacula 9.6.5 rpms > packages. Got the Fedora31 9.6.5 client installed today with no issues (except moving the config file from /etc/bacula to /opt/bacula/etc). Thanks again Davide! -- Phil Stracchino Babylon Communications ph...@caerllewys.net p...@co.ordinate.org Landline: +1.603.293.8485 Mobile: +1.603.998.6958 ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] bacula - optimize storage for cloud sync
Hello Ziga, Yes, you might be able to do what you want using a "debug" feature of the Bacula Cloud driver. It is not very well documented, but there is one section "3 File Driver for the Cloud" in the "Bacula Cloud Backup" that mentions it. Basically instead of using the "S3" driver in the Cloud resource of your Storage Daemon, you use "File" and the HostName becomes the path where the Cloud volumes (directories + parts) will be written. For example, I use the following for writing to disk instead of an S3 cloud. Cloud { Name = DummyCloud Driver = "File" HostName = "/home/kern/bacula/k/regress/tmp/cloud" BucketName = "DummyBucket" AccessKey = "DummyAccessKey" SecretKey = "DummySecretKey" Protocol = HTTPS UriStyle = VirtualHost } The Device resource looks like: Device { Name = FileStorage1 Media Type = File1 Archive Device = /home/kern/bacula/k/regress/tmp LabelMedia = yes; # lets Bacula label unlabelled media Random Access = Yes; AutomaticMount = yes; # when device opened, read it RemovableMedia = no; AlwaysOpen = no; Device Type = Cloud Cloud = DummyCloud } I know the code runs and produces correct output, but I am not sure how it will work in your environment. If it works, great. If it doesn't work, for the near future, unfortunately I cannot provide support, but at some point (probably 3-6 months) the project may support this feature. Note: the next version of Bacula coming in a few months will have a good number of new features and improvements to the Cloud driver (more bconsole commands for example). Good luck, and best regards, Kern PS: If it does work for you, I would appreciate it if you would document it and publish it on the bacula-users list so others can use the Oracle cloud. On 7/8/20 3:41 PM, Žiga Žvan wrote: Hi Mr. Kern, My question was a bit different. I have noticed that Oracle S3 is not compatible, therefore I have implemented Oracle Storage gateway (a docker image that uses local filesystem as a cache and moves the data automatically to oracle cloud). I have this filesystem mounted (nfsv4) on bacula server and I am able to backup data to this storage (and hence in cloud). I have around 1 TB data daily and I'm a bit concerned about the bandwidth. It will take app. 4 hours to sync to the cloud and I need to count in the future growth. As long as bacula writes data to one file/volume, where it stores full and incremental backups, this is not optimal for the cloud (the file will change and all the data will upload each day). I have noticed that bacula stores data differently in the cloud configuration. Volume is not a file, but a folder with fileparts. This would be better for me, because only some fileparts would change and move to the cloud via Storage gateway. So the question is: Can I configure bacula-sd to store data in fileparts, without actual cloud sync? Is this possible? I have tried several configurations of a bacula-sd device with no luck. Should I configure some dummy cloud resource? Kind regards, Ziga Zvan On 07/07/2020 14:40, Kern Sibbald wrote: Hello, Oracle S3 is not compatible with Amazon S3 or at least with the libs3 that we use to interface to AWS and other compatible S3 cloud offerings. Yes, Bacula Enterprise has a separate Oracle cloud driver that they wrote. There are no plans at the moment to backport it to the community version. Best regards, Kern On 7/7/20 8:43 AM, Žiga Žvan wrote: Dear all, I'm testing communty version of bacula in order to change backup sw for app. 100 virtual and physical hosts. I would like to move all the data to local storage and then move them to public cloud (Oracle Object storage). I believe that community version of the software suites our needs. I have installed: -version 9.6.5 of bacula on centos 7 computer -oracle storage gateway (similar to aws SG - it moves data to object storage and exposes it localy as nfsv4; for bacula this is backup destination). I have read this two documents regarding bacula and
Re: [Bacula-users] Community binaries on Fedora 31: libreadline.so.7 not available
On 2020-07-08 02:54, Davide Franco wrote: > Hello Phil, > > You dont need a dev environment as I have uploaded Bacula 9.6.5 rpms > packages. > > Fedora 32 packages will come soon. > > Best regards > > Davide Thanks Davide. -- Phil Stracchino Babylon Communications ph...@caerllewys.net p...@co.ordinate.org Landline: +1.603.293.8485 Mobile: +1.603.998.6958 ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] bacula - optimize storage for cloud sync
Hi Mr. Kern, My question was a bit different. I have noticed that Oracle S3 is not compatible, therefore I have implemented Oracle Storage gateway (a docker image that uses local filesystem as a cache and moves the data automatically to oracle cloud). I have this filesystem mounted (nfsv4) on bacula server and I am able to backup data to this storage (and hence in cloud). I have around 1 TB data daily and I'm a bit concerned about the bandwidth. It will take app. 4 hours to sync to the cloud and I need to count in the future growth. As long as bacula writes data to one file/volume, where it stores full and incremental backups, this is not optimal for the cloud (the file will change and all the data will upload each day). I have noticed that bacula stores data differently in the cloud configuration. Volume is not a file, but a folder with fileparts. This would be better for me, because only some fileparts would change and move to the cloud via Storage gateway. So the question is: Can I configure bacula-sd to store data in fileparts, without actual cloud sync? Is this possible? I have tried several configurations of a bacula-sd device with no luck. Should I configure some dummy cloud resource? Kind regards, Ziga Zvan On 07/07/2020 14:40, Kern Sibbald wrote: Hello, Oracle S3 is not compatible with Amazon S3 or at least with the libs3 that we use to interface to AWS and other compatible S3 cloud offerings. Yes, Bacula Enterprise has a separate Oracle cloud driver that they wrote. There are no plans at the moment to backport it to the community version. Best regards, Kern On 7/7/20 8:43 AM, Žiga Žvan wrote: Dear all, I'm testing communty version of bacula in order to change backup sw for app. 100 virtual and physical hosts. I would like to move all the data to local storage and then move them to public cloud (Oracle Object storage). I believe that community version of the software suites our needs. I have installed: -version 9.6.5 of bacula on centos 7 computer -oracle storage gateway (similar to aws SG - it moves data to object storage and exposes it localy as nfsv4; for bacula this is backup destination). I have read this two documents regarding bacula and cloud https://blog.bacula.org/whitepapers/CloudBackup.pdf https://blog.bacula.org/whitepapers/ObjectStorage.pdf It is mentioned it the document above, that Oracle Object storage is not supported at the moment. Is it possible to *configure* bacula Storage device in a way that uses *Cloud media format* (directory with file parts as a volume, instead of a single file as a volume) *without actual cloud sync* (Storage Gateway does this in my case)? I am experimenting with variations of the definition bellow, but I am unable to solve this issue for now (it tries to initialize cloud plugin or it writes to a file, instead of a directory). Device { Name = FSOciCloudStandard # Device type = Cloud Device type = File # Cloud = OracleViaStorageGateway Maximum Part Size = 100 MB # Media Type = File Media Type = CloudType Archive Device = /mnt/baculatest_standard/backup LabelMedia = yes; # lets Bacula label unlabeled media Random Access = Yes; AutomaticMount = yes; # when device opened, read it RemovableMedia = no; AlwaysOpen = no; } Is there any plan to support oracle object storage in near future? It has S3 compatible API and bacula enterprise supports it... Kind regards, Ziga Zvan ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users