View Issue Details
ID | Project | Category | View Status | Date Submitted | Last Update |
---|---|---|---|---|---|
0001271 | bareos-core | storage daemon | public | 2020-09-08 23:16 | 2021-09-28 15:27 |
Reporter | kabassanov | Assigned To | bruno-at-bareos | ||
Priority | high | Severity | major | Reproducibility | always |
Status | closed | Resolution | unable to reproduce | ||
Platform | Linux | OS | Debian | OS Version | 10 |
Product Version | 19.2.8 | ||||
Summary | 0001271: Btape speed tests with Bareos block structure have half of the Bacula throughput | ||||
Description | Hello, Some months ago I migrated a backup server from Bacula to Bareos. I've just remarked that with btape tests (and real backups) write throughtput is decreased by about 50% when bareos block structure is used. So I wonder if there is something wrong there... Like a bareos block structure that overlaps and needs 2 blocks while sending data to the drive? Bacula: ---------- btape: btape.c:1062-0 Test with zero data, should give the maximum throughput. btape: btape.c:911-0 Begin writing 3 files of 3.221 GB with raw blocks of 1048576 bytes. btape: btape.c:385-0 Total Volume bytes=9.663 GB. Total Write rate = 508.6 MB/s +++ btape: btape.c:1074-0 Test with random data, should give the minimum throughput. btape: btape.c:911-0 Begin writing 3 files of 3.221 GB with raw blocks of 1048576 bytes. btape: btape.c:385-0 Total Volume bytes=9.663 GB. Total Write rate = 214.7 MB/s +++ btape: btape.c:1088-0 Test with zero data and bacula block structure. btape: btape.c:966-0 Begin writing 3 files of 3.221 GB with blocks of 1048576 bytes. btape: btape.c:385-0 Total Volume bytes=9.663 GB. Total Write rate = 536.8 MB/s btape: btape.c:1100-0 Test with random data, should give the minimum throughput. btape: btape.c:966-0 Begin writing 3 files of 3.221 GB with blocks of 1048576 bytes. btape: btape.c:385-0 Total Volume bytes=9.663 GB. Total Write rate = 214.7 MB/s Bareos: btape: stored/btape.cc:1026-0 Test with zero data, should give the maximum throughput. btape: stored/btape.cc:887-0 Begin writing 3 files of 3.221 GB with raw blocks of 1048576 bytes. btape: stored/btape.cc:396-0 Total Volume bytes=9.663 GB. Total Write rate = 536.8 MB/s +++ btape: stored/btape.cc:1038-0 Test with random data, should give the minimum throughput. btape: stored/btape.cc:887-0 Begin writing 3 files of 3.221 GB with raw blocks of 1048576 bytes. btape: stored/btape.cc:396-0 Total Volume bytes=9.663 GB. Total Write rate = 214.7 MB/s +++ btape: stored/btape.cc:1052-0 Test with zero data and bareos block structure. btape: stored/btape.cc:942-0 Begin writing 3 files of 3.221 GB with blocks of 1048576 bytes. btape: stored/btape.cc:396-0 Total Volume bytes=9.663 GB. Total Write rate = 292.8 MB/s +++ btape: stored/btape.cc:1064-0 Test with random data, should give the minimum throughput. btape: stored/btape.cc:942-0 Begin writing 3 files of 3.221 GB with blocks of 1048576 bytes. btape: stored/btape.cc:396-0 Total Volume bytes=9.663 GB. Total Write rate = 108.5 MB/s | ||||
Tags | No tags attached. | ||||
Hello, thanks for the report, would you mind to give a bit more information ? bacula version used and edition Manufacturer name of tape output of sudo lshw -json output of uname -a Regards |
|
Hi, Bacula 9.4.2 (04Feb19) Tapes: Fujifilm LTO Ultrium 7 - 6 TB /15 TB Attached LTO related part of lshw. If more needed, please contact me in private. Linux "hostname" 5.10.0-1-amd64 0000001 SMP Debian 5.10.4-1 (2020-12-31) x86_64 GNU/Linux (probably kernel has been updated since bacula btape tests in August 2019...) Regards lshw-lto-related.txt (5,167 bytes)
{ "id" : "pci:3", "class" : "bridge", "claimed" : true, "handle" : "PCIBUS:0000:83", "description" : "PCI bridge", "product" : "Xeon E7 v4/Xeon E5 v4/Xeon E3 v4/Xeon D PCI Express Root Port 2", "vendor" : "Intel Corporation", "physid" : "2", "businfo" : "pci@0000:80:02.0", "version" : "01", "width" : 32, "clock" : 33000000, "configuration" : { "driver" : "pcieport" }, "capabilities" : { "pci" : true, "msi" : "Message Signalled Interrupts", "pciexpress" : "PCI Express", "pm" : "Power Management", "normal_decode" : true, "bus_master" : "bus mastering", "cap_list" : "PCI capabilities listing" } { "id" : "fiber:0", "class" : "bus", "claimed" : true, "handle" : "PCI:0000:83:00.0", "description" : "Fibre Channel", "product" : "ISP2532-based 8Gb Fibre Channel to PCI Express HBA", "vendor" : "QLogic Corp.", "physid" : "0", "businfo" : "pci@0000:83:00.0", "logicalname" : "scsi1", "version" : "02", "width" : 64, "clock" : 33000000, "configuration" : { "driver" : "qla2xxx", "latency" : "0" }, "capabilities" : { "pm" : "Power Management", "pciexpress" : "PCI Express", "msi" : "Message Signalled Interrupts", "vpd" : "Vital Product Data", "msix" : "MSI-X", "bus_master" : "bus mastering", "cap_list" : "PCI capabilities listing", "rom" : "extension ROM" } { "id" : "tape", "class" : "tape", "handle" : "SCSI:01:00:00:00", "description" : "SCSI Tape", "product" : "ULTRIUM-HH7", "vendor" : "IBM", "physid" : "0.0.0", "businfo" : "scsi@1:0.0.0", "version" : "M571", "serial" : "XXX", "configuration" : { "ansiversion" : "6" }, "capabilities" : { "removable" : "support is removable" } }, { "id" : "changer", "class" : "generic", "handle" : "SCSI:01:00:00:01", "description" : "SCSI Medium Changer", "product" : "FlexStor II", "vendor" : "BDT", "physid" : "0.0.1", "businfo" : "scsi@1:0.0.1", "version" : "5.60", "serial" : "XXX", "configuration" : { "ansiversion" : "5" }, "capabilities" : { "removable" : "support is removable" } }, }, { "id" : "fiber:1", "class" : "bus", "claimed" : true, "handle" : "PCI:0000:83:00.1", "description" : "Fibre Channel", "product" : "ISP2532-based 8Gb Fibre Channel to PCI Express HBA", "vendor" : "QLogic Corp.", "physid" : "0.1", "businfo" : "pci@0000:83:00.1", "logicalname" : "scsi13", "version" : "02", "width" : 64, "clock" : 33000000, "configuration" : { "driver" : "qla2xxx", "latency" : "0" }, "capabilities" : { "pm" : "Power Management", "pciexpress" : "PCI Express", "msi" : "Message Signalled Interrupts", "vpd" : "Vital Product Data", "msix" : "MSI-X", "bus_master" : "bus mastering", "cap_list" : "PCI capabilities listing", "rom" : "extension ROM" } { "id" : "tape", "class" : "tape", "handle" : "SCSI:13:00:00:00", "description" : "SCSI Tape", "product" : "ULTRIUM-HH7", "vendor" : "IBM", "physid" : "0.0.0", "businfo" : "scsi@13:0.0.0", "version" : "M571", "serial" : "XXX", "configuration" : { "ansiversion" : "6" }, "capabilities" : { "removable" : "support is removable" } }, }, } |
|
Hello we wonder if you can share your sd configuration (/etc/bareos/bareos-dir.d/storage/* + /etc/bareos/bareos-sd.d/*) Don't forget to remove the Password line. Thanks. |
|
Here it is... cat /etc/bareos/bareos-sd.d/device/Autochanger.conf # # An autochanger device with two drives # Autochanger { Name = LTO-7-Autochanger Device = Drive-1, Drive-2 Changer Device = /dev/changer Changer Command = "/etc/bareos/custom-configs/scripts/mtx-changer %c %o %S %a %d" Description = "Tape device library. A connecting Director must have the same Name and MediaType." } cat /etc/bareos/bareos-sd.d/device/Drive-1.conf # A Linux or Solaris LTO-7 tape drive # Device { Name = Drive-1 Drive Index = 0 Media Type = LTO-7 Archive Device = /dev/nst/tape0 AutoChanger = yes LabelMedia = no; AutomaticMount = yes; # when device opened, read it AlwaysOpen = yes; RemovableMedia = yes; RandomAccess = no; Minimum Block Size = 1048576 Maximum Block Size = 1048576 Label Block Size = 1048576 Maximum Concurrent Jobs = 1 Spool Directory = /bareos-spool Maximum Spool Size = 4398046511104 } cat /etc/bareos/bareos-sd.d/device/Drive-2.conf Device { Name = Drive-2 Drive Index = 1 Media Type = LTO-7 Archive Device = /dev/nst/tape1 AutoChanger = yes LabelMedia = no; AutomaticMount = yes; # when device opened, read it AlwaysOpen = yes; RemovableMedia = yes; RandomAccess = no; Minimum Block Size = 1048576 Maximum Block Size = 1048576 Label Block Size = 1048576 Maximum Concurrent Jobs = 1 Spool Directory = /bareos-spool Maximum Spool Size = 4398046511104 } Storage { Name = XXXXX Maximum Concurrent Jobs = 100 SDAddress = XXXXX SDPort = 9103 # Director's port WorkingDirectory = "/var/lib/bareos" Pid Directory = "/run/bareos" TLS Enable = No } cat /etc/bareos/bareos-dir.d/storage/Changer.conf # ----------------------------------------------- # Definition of LTO-7 tape changer in Autochanger device Storage { Name = Changer # Do not use "localhost" here Address = XXXXX # N.B. Use a fully qualified name here SDPort = 9103 Password = "xxxxxx" Device = LTO-7-Autochanger Auto Changer = yes Media Type = LTO-7 # must be same as MediaType in Storage daemon Maximum Concurrent Jobs = 10 TLS Enable = No } Regards. |
|
Hello, we tried to reproduce the reported behavior but didn't success. It was not possible to reproduce the mis-behaviour reported by you, we tried on VM with using mhvtl with actual code. We have previously some reports of tests with LTO4 drives (code was at 12.4 and 13.2) with no differences too. I was able to reproduce one time by using device with compression off for bareos, and compression on for bacula. Then bareos is 100% slower than bacula, but that's trying to compare apples and bananas. Would it be possible that compression was disabled when you tried Bareos ? Methodology : btape is writing 3 times the file of 3.221 GB and return an average. The test is run 3 times, and then the average of the 3 resulting averages is summed in the table below. ``` |-------------|-----------|-------------|-------------------------| |Run | zero data | random data | zero+block | rand+block | |-------------|-----------|-------------|-------------------------| |bacula 64 | 327.6 | 233.6 | 124.0 | 226.7 | |bacula 256 | 328.1 | 297.4 | 193.5 | 214.4 | |bacula 512 | 371.6 | 297.4 | 264.8 | 233.7 | |_Average_ bac| 342.4 | 276.1 | 194.1 | 224.9 | |-------------|-----------|-------------|-------------------------| |bareos 64 | 307.7 | 268.6 | 128.4 | 178.6 | |bareos 256 | 268.5 | 292.8 | 195.2 | 167.0 | |bareos 512 | 331.5 | 272.7 | 261.1 | 210.0 | |_Average_ bar| 302.6 | 278.0 | 192.9 | 222.8 | |-------------|-----------|-------------|-------------------------| |_FactorXslow_| 1.13 | 0.99 | 1.00 | 1.00 | |-------------|-----------|-------------|-------------------------| ``` **Note** mhvtl doesn't work with block of 1048576. I will close this report as non-fix (we don't know what to fix, as we can't reproduce it actually). But please feel free to reopen it if you have any other clue or evidence to add. Best regards. |
|
see previous note | |
Date Modified | Username | Field | Change |
---|---|---|---|
2020-09-08 23:16 | kabassanov | New Issue | |
2021-09-07 09:27 | bruno-at-bareos | Note Added: 0004251 | |
2021-09-11 17:14 | kabassanov | File Added: lshw-lto-related.txt | |
2021-09-11 17:14 | kabassanov | Note Added: 0004262 | |
2021-09-21 13:30 | bruno-at-bareos | Note Added: 0004269 | |
2021-09-21 14:14 | kabassanov | Note Added: 0004270 | |
2021-09-28 15:26 | bruno-at-bareos | Note Added: 0004281 | |
2021-09-28 15:27 | bruno-at-bareos | Assigned To | => bruno-at-bareos |
2021-09-28 15:27 | bruno-at-bareos | Status | new => assigned |
2021-09-28 15:27 | bruno-at-bareos | Status | assigned => closed |
2021-09-28 15:27 | bruno-at-bareos | Resolution | open => unable to reproduce |
2021-09-28 15:27 | bruno-at-bareos | Note Added: 0004282 |