Re: [gpfsug-discuss] subblock sanity check in 5.0
Joey, The subblocks-per-full-block value cannot be specified when the file system is created, but is rather computed automatically by GPFS. In file systems with format older than 5.0, the value is fixed at 32. For file systems with format 5.0.0 or later, the value is computed based on the block size. See manpage for mmcrfs, in table where the -B BlockSize option is explained. (Table 1. Block sizes and subblock sizes) . Say, for the default (in 5.0+) 4MB block size, the subblock size is 8KB. The minimum "practical" subblock size is 4KB, to keep 4KB-alignment to accommodate 4KN devices. Felipe Felipe Knop k...@us.ibm.com GPFS Development and Security IBM Systems IBM Building 008 2455 South Rd, Poughkeepsie, NY 12601 (845) 433-9314 T/L 293-9314 From: Joseph Mendoza To: gpfsug main discussion list Date: 06/25/2018 08:59 PM Subject:[gpfsug-discuss] subblock sanity check in 5.0 Sent by:gpfsug-discuss-boun...@spectrumscale.org Quick question, anyone know why GPFS wouldn't respect the default for the subblocks-per-full-block parameter when creating a new filesystem? I'd expect it to be set to 512 for an 8MB block size but my guess is that also specifying a metadata-block-size is interfering with it (by being too small). This was a parameter recommended by the vendor for a 4.2 installation with metadata on dedicated SSDs in the system pool, any best practices for 5.0? I'm guessing I'd have to bump it up to at least 4MB to get 512 subblocks for both pools. fs1 created with: # mmcrfs fs1 -F fs1_ALL -A no -B 8M -i 4096 -m 2 -M 2 -r 1 -R 2 -j cluster -n 9000 --metadata-block-size 512K --perfileset-quota --filesetdf -S relatime -Q yes --inode-limit 2000:1000 -T /gpfs/fs1 # mmlsfs fs1 flag value description --- --- -f 8192 Minimum fragment (subblock) size in bytes (system pool) 131072 Minimum fragment (subblock) size in bytes (other pools) -i 4096 Inode size in bytes -I 32768 Indirect block size in bytes -B 524288 Block size (system pool) 8388608 Block size (other pools) -V 19.01 (5.0.1.0) File system version --subblocks-per-full-block 64 Number of subblocks per full block -P system;DATA Disk storage pools in file system Thanks! --Joey Mendoza NCAR ___ gpfsug-discuss mailing list gpfsug-discuss at spectrumscale.org http://gpfsug.org/mailman/listinfo/gpfsug-discuss ___ gpfsug-discuss mailing list gpfsug-discuss at spectrumscale.org http://gpfsug.org/mailman/listinfo/gpfsug-discuss
[gpfsug-discuss] subblock sanity check in 5.0
Quick question, anyone know why GPFS wouldn't respect the default for the subblocks-per-full-block parameter when creating a new filesystem? I'd expect it to be set to 512 for an 8MB block size but my guess is that also specifying a metadata-block-size is interfering with it (by being too small). This was a parameter recommended by the vendor for a 4.2 installation with metadata on dedicated SSDs in the system pool, any best practices for 5.0? I'm guessing I'd have to bump it up to at least 4MB to get 512 subblocks for both pools. fs1 created with: # mmcrfs fs1 -F fs1_ALL -A no -B 8M -i 4096 -m 2 -M 2 -r 1 -R 2 -j cluster -n 9000 --metadata-block-size 512K --perfileset-quota --filesetdf -S relatime -Q yes --inode-limit 2000:1000 -T /gpfs/fs1 # mmlsfs fs1 flag value description --- --- -f 8192 Minimum fragment (subblock) size in bytes (system pool) 131072 Minimum fragment (subblock) size in bytes (other pools) -i 4096 Inode size in bytes -I 32768 Indirect block size in bytes -B 524288 Block size (system pool) 8388608 Block size (other pools) -V 19.01 (5.0.1.0) File system version --subblocks-per-full-block 64 Number of subblocks per full block -P system;DATA Disk storage pools in file system Thanks! --Joey Mendoza NCAR ___ gpfsug-discuss mailing list gpfsug-discuss at spectrumscale.org http://gpfsug.org/mailman/listinfo/gpfsug-discuss
[gpfsug-discuss] mmchconfig subnets
Hi, I'm hoping somebody has insights into how the subnets option actually works. I've read the docs a dozen times and I want to make sure I understand before I take my production cluster down to make the changes. On the current cluster the daemon addresses are on a gpfs private network and the admin addresses are on a public network. I'm changing so both daemon and admin are public and the subnets option is used to utilize the private network. This is to facilitate remote mounts to an independent cluster. The confusing factor in my case, not covered in the docs, is that the gpfs private network is subnetted and static routes are used to reach them. That is, there are three private networks, one for each datacenter and the cluster nodes daemon interfaces are spread between the three. 172.16.141.32/27 172.16.141.24/29 172.16.141.128/27 A router connects these three networks but are otherwise 100% private. For my mmchconfig subnets command should I use this? mmchconfig subnets="172.16.141.24 172.16.141.32 172.16.141.128" Where I get confused is that I'm trying to reason through how Spectrum Scale is utilizing the subnets setting to decide if this will have the desired result on my cluster. If I change the node addresses to their public addresses, ie the private addresses are not explicitly configured in Scale, then how are the private addresses discovered? Does each node use the subnets option to identify that it has a private address and then dynamically shares that with the cluster? Thanks in advance for your clarifying comments. -Eric -- Eric Horst University of Washington ___ gpfsug-discuss mailing list gpfsug-discuss at spectrumscale.org http://gpfsug.org/mailman/listinfo/gpfsug-discuss
Re: [gpfsug-discuss] mmapplypolicy on nested filesets ...
It took a while before I could get back to this issue, but I want to confirm that Marc's suggestions worked line a charm, and did exactly what I hoped for: * remove any FOR FILESET(...) specifications * mmapplypolicy /path/to/the/root/directory/of/the/independent-fileset-you-wish-to-scan ... --scope inodespace -P your-policy-rules-file ... I didn't have to do anything else, but exclude a few filesets from the scan. Thanks Jaime Quoting "Marc A Kaplan" : I suggest you remove any FOR FILESET(...) specifications from your rules and then run mmapplypolicy /path/to/the/root/directory/of/the/independent-fileset-you-wish-to-scan ... --scope inodespace -P your-policy-rules-file ... See also the (RTFineM) for the --scope option and the Directory argument of the mmapplypolicy command. That is the best, most efficient way to scan all the files that are in a particular inode-space. Also, you must have all filesets of interest "linked" and the file system must be mounted. Notice that "independent" means that the fileset name is used to denote both a fileset and an inode-space, where said inode-space contains the fileset of that name and possibly other "dependent" filesets... IF one wished to search the entire file system for files within several different filesets, one could use rules with FOR FILESET('fileset1','fileset2','and-so-on') Or even more flexibly WHERE FILESET_NAME LIKE 'sql-like-pattern-with-%s-and-maybe-_s' Or even more powerfully WHERE regex(FILESET_NAME, 'extended-regular-.*-expression') From: "Jaime Pinto" To: "gpfsug main discussion list" Date: 04/18/2018 01:00 PM Subject:[gpfsug-discuss] mmapplypolicy on nested filesets ... Sent by:gpfsug-discuss-boun...@spectrumscale.org A few months ago I asked about limits and dynamics of traversing depended .vs independent filesets on this forum. I used the information provided to make decisions and setup our new DSS based gpfs storage system. Now I have a problem I couldn't' yet figure out how to make it work: 'project' and 'scratch' are top *independent* filesets of the same file system. 'proj1', 'proj2' are dependent filesets nested under 'project' 'scra1', 'scra2' are dependent filesets nested under 'scratch' I would like to run a purging policy on all contents under 'scratch' (which includes 'scra1', 'scra2'), and TSM backup policies on all contents under 'project' (which includes 'proj1', 'proj2'). HOWEVER: When I run the purging policy on the whole gpfs device (with both 'project' and 'scratch' filesets) * if I use FOR FILESET('scratch') on the list rules, the 'scra1' and 'scra2' filesets under scratch are excluded (totally unexpected) * if I use FOR FILESET('scra1') I get error that scra1 is dependent fileset (Ok, that is expected) * if I use /*FOR FILESET('scratch')*/, all contents under 'project', 'proj1', 'proj2' are traversed as well, and I don't want that (it takes too much time) * if I use /*FOR FILESET('scratch')*/, and instead of the whole device I apply the policy to the /scratch mount point only, the policy still traverses all the content of 'project', 'proj1', 'proj2', which I don't want. (again, totally unexpected) QUESTION: How can I craft the syntax of the mmapplypolicy in combination with the RULE filters, so that I can traverse all the contents under the 'scratch' independent fileset, including the nested dependent filesets 'scra1','scra2', and NOT traverse the other independent filesets at all (since this takes too much time)? Thanks Jaime PS: FOR FILESET('scra*') does not work. TELL US ABOUT YOUR SUCCESS STORIES https://urldefense.proofpoint.com/v2/url?u=http-3A__www.scinethpc.ca_testimonials=DwICAg=jf_iaSHvJObTbx-siA1ZOg=cvpnBBH0j41aQy0RPiG2xRL_M8mTc1izuQD3_PmtjZ8=y0aRzkzp0QA9QR8eh3XtN6PETqWYDCNvItdihzdueTE=IpwHlr0YNr7rgV7gI8Y2sxIELLIwA15KK4nBnv9BYWk= --- Jaime Pinto - Storage Analyst SciNet HPC Consortium - Compute/Calcul Canada www.scinet.utoronto.ca - www.computecanada.ca University of Toronto 661 University Ave. (MaRS), Suite 1140 Toronto, ON, M5G1M1 P: 416-978-2755 C: 416-505-1477 This message was sent using IMP at SciNet Consortium, University of Toronto. ___ gpfsug-discuss mailing list gpfsug-discuss at spectrumscale.org https://urldefense.proofpoint.com/v2/url?u=http-3A__gpfsug.org_mailman_listinfo_gpfsug-2Ddiscuss=DwICAg=jf_iaSHvJObTbx-siA1ZOg=cvpnBBH0j41aQy0RPiG2xRL_M8mTc1izuQD3_PmtjZ8=y0aRzkzp0QA9QR8eh3XtN6PETqWYDCNvItdihzdueTE=aff0vMJkKd-Z3pw3-jckmI3ejqXh8aSr8rxkKf3OGdk= TELL US ABOUT YOUR SUCCESS STORIES http://www.scinethpc.ca/testimonials --- Jaime Pinto - Storage Analyst SciNet HPC Consortium -
Re: [gpfsug-discuss] mmbackup issue
Hallo All, here the requirement for enhancement of mmbackup. http://www.ibm.com/developerworks/rfe/execute?use_case=viewRfe_ID=121687 Please vote. Renar Grunenberg Abteilung Informatik – Betrieb HUK-COBURG Bahnhofsplatz 96444 Coburg Telefon: 09561 96-44110 Telefax: 09561 96-44104 E-Mail: renar.grunenb...@huk-coburg.de Internet: www.huk.de === HUK-COBURG Haftpflicht-Unterstützungs-Kasse kraftfahrender Beamter Deutschlands a. G. in Coburg Reg.-Gericht Coburg HRB 100; St.-Nr. 9212/101/00021 Sitz der Gesellschaft: Bahnhofsplatz, 96444 Coburg Vorsitzender des Aufsichtsrats: Prof. Dr. Heinrich R. Schradin. Vorstand: Klaus-Jürgen Heitmann (Sprecher), Stefan Gronbach, Dr. Hans Olav Herøy, Dr. Jörg Rheinländer (stv.), Sarah Rössler, Daniel Thomas. === Diese Nachricht enthält vertrauliche und/oder rechtlich geschützte Informationen. Wenn Sie nicht der richtige Adressat sind oder diese Nachricht irrtümlich erhalten haben, informieren Sie bitte sofort den Absender und vernichten Sie diese Nachricht. Das unerlaubte Kopieren sowie die unbefugte Weitergabe dieser Nachricht ist nicht gestattet. This information may contain confidential and/or privileged information. If you are not the intended recipient (or have received this information in error) please notify the sender immediately and destroy this information. Any unauthorized copying, disclosure or distribution of the material in this information is strictly forbidden. === -Ursprüngliche Nachricht- Von: gpfsug-discuss-boun...@spectrumscale.org [mailto:gpfsug-discuss-boun...@spectrumscale.org] Im Auftrag von Jonathan Buzzard Gesendet: Donnerstag, 21. Juni 2018 09:33 An: gpfsug-discuss@spectrumscale.org Betreff: Re: [gpfsug-discuss] mmbackup issue On 20/06/18 17:00, Grunenberg, Renar wrote: > Hallo Valdis, first thanks for the explanation we understand that, > but this problem generate only 2 Version at tsm server for the same > file, in the same directory. This mean that mmbackup and the > .shadow... has no possibility to have for the same file in the same > directory more then 2 backup versions with tsm. The native ba-client > manage this. (Here are there already different inode numbers > existent.) But at TSM-Server side the file that are selected at 'ba > incr' are merged to the right filespace and will be binded to the > mcclass >2 Version exist. > I think what you are saying is that mmbackup is only keeping two versions of the file in the backup, the current version and a single previous version. Normally in TSM you can control how many previous versions of the file that you can keep, for both active and inactive (aka deleted). You can also define how long these version are kept for. It sounds like you are saying that mmbackup is ignoring the policy that you have set for this in TSM (q copy) and doing it's own thing? JAB. -- Jonathan A. Buzzard Tel: +44141-5483420 HPC System Administrator, ARCHIE-WeSt. University of Strathclyde, John Anderson Building, Glasgow. G4 0NG ___ gpfsug-discuss mailing list gpfsug-discuss at spectrumscale.org http://gpfsug.org/mailman/listinfo/gpfsug-discuss ___ gpfsug-discuss mailing list gpfsug-discuss at spectrumscale.org http://gpfsug.org/mailman/listinfo/gpfsug-discuss