Re: I/O delays
Just incase anyone was interested in the resolution to this problem, it appears that the poor drive performance was caused by vibrations from the chassis fans. After replacing about every component in this system (hard drive, motherboard, SATA cable, memory, and CPU's) to no avail, we discovered that our chassis had a mix of 6.8W and 2.8W fans. When we corrected this so that all five chassis fans were of the 2.8W variety, our disk performance problems appeared to disappear. For what it's worth, a quick search produces a few reports of vibrations degrading disk performance. For example: http://www.dtc.umn.edu/publications/reports/2005_08.pdf http://etbe.coker.com.au/2009/04/22/vibration-strange-sata-performance/ Finally to quantify the improvement in disk speed, here are the results (showing the average and range) of each test from three runs of bonnie++: bonnie++ -d /mnt/scratch/test_dab66 -n 0 -s 98304 -f ORIGINAL FAN CONFIGURATION (two 2.8W and three 6.8W) Writes AVG = 5348; 4327 - 6369 Rewrites AVG = 2489.5; 2284 - 2695 Reads AVG = 9575.5; 7778 - 11373 Random Seeks AVG = 68.4; 65.3 - 71.5 CORRECT FAN CONFIGURATION (five 2.8W) Writes AVG = 53901.333; 43143 - 60430 Rewrites AVG = 27684; 20178 - 31942 Reads AVG = 75541.333; 65704 - 80717 Random Seeks AVG = 181.333; 117.9 - 218.6 Many thanks for everyone's time and help. Sincerely, Devin
Re: I/O delays
Hi Stijn, Thank you very much for your reply. On Sep 9, 2010, at 2:13 AM, Stijn De Weirdt wrote: > so there is only 1 disk in your system according to dmesg (a 500GB SATA > disk) and dd reports speeds > 1GB/s. you are either the lucky owner of a > wonder SATA disk or you are measuring linux caching ;) Yes, indeed =). > when you run the tests, can you also run > watch grep Dirty /proc/meminfo > and check if it starts increasing, what the maximum is and when it > starts decreasing. It appears to start decreasing at 1GB. With /proc/sys/vm/dirty_ratio set to 40, the maximum I've seen is 1030708 kB. With it set to 2, the max I've seen is 991704 kB. > if you have default SL settings /proc/sys/vm/dirty_ratio is 40, so > that's 19GB of dirty memory allowed. with your dd you won't reach that, > but default /proc/sys/vm/dirty_expire_centisecs is 3000, so after 30 > seconds, the system will start to write the dirty memory away, which is > probably when the "reduced performance" starts to hit you. > > you can safely lower /proc/sys/vm/dirty_ratio to 2 (which is still > almost 1GB of dirty memory) to get a more stable performance (i think > recent kernel do this automated based on total size of memory) > > and if you want to measure your disk performance, use iozone/bonnie/IOR > with sync options (or add 'sync' to you timing stats) The system is currently in use and I haven't yet had a chance to run any real disk benchmarks. For what it's worth, here's the dd output with a subsequent "time sync". During the sync, the system as a whole becomes sluggish (as we only have a single drive used both for the OS and user data) and it takes several seconds for commands like "top" to return. We also see comparable behavior when using ext3 or XFS. [da...@lnx4103 ~]% time dd if=/dev/zero of=/mnt/scratch/test bs=1M count=1K ; time sync 1024+0 records in 1024+0 records out 1073741824 bytes (1.1 GB) copied, 0.928643 seconds, 1.2 GB/s real0m1.077s user0m0.000s sys 0m1.068s real1m35.341s user0m0.000s sys 0m0.182s > btw, try dstat (sort of iostat and vmstat combined, with colors ;) Yes, dstat is very nice! We will let you know whether or not things improve after replacing the disk. Thanks again, Devin
Re: I/O delays
hi devin, so there is only 1 disk in your system according to dmesg (a 500GB SATA disk) and dd reports speeds > 1GB/s. you are either the lucky owner of a wonder SATA disk or you are measuring linux caching ;) when you run the tests, can you also run watch grep Dirty /proc/meminfo and check if it starts increasing, what the maximum is and when it starts decreasing. if you have default SL settings /proc/sys/vm/dirty_ratio is 40, so that's 19GB of dirty memory allowed. with your dd you won't reach that, but default /proc/sys/vm/dirty_expire_centisecs is 3000, so after 30 seconds, the system will start to write the dirty memory away, which is probably when the "reduced performance" starts to hit you. you can safely lower /proc/sys/vm/dirty_ratio to 2 (which is still almost 1GB of dirty memory) to get a more stable performance (i think recent kernel do this automated based on total size of memory) and if you want to measure your disk performance, use iozone/bonnie/IOR with sync options (or add 'sync' to you timing stats) btw, try dstat (sort of iostat and vmstat combined, with colors ;) stijn > Hi Konstantin, > > Thank you for the reply. Here is the output of smartctl. We will see if a > different drive model helps. > > Sincerely, > Devin > > [r...@lnx4103 ~]# smartctl -a /dev/sda > smartctl version 5.38 [x86_64-redhat-linux-gnu] Copyright (C) 2002-8 Bruce > Allen > Home page is http://smartmontools.sourceforge.net/ > > === START OF INFORMATION SECTION === > Device Model: WDC WD5002ABYS-02B1B0 > Serial Number:WD-WCASYDXX > Firmware Version: 02.03B03 > User Capacity:500,107,862,016 bytes > Device is:Not in smartctl database [for details use: -P showall] > ATA Version is: 8 > ATA Standard is: Exact ATA specification draft version not indicated > Local Time is:Wed Sep 8 23:34:17 2010 EDT > SMART support is: Available - device has SMART capability. > SMART support is: Enabled > > === START OF READ SMART DATA SECTION === > SMART overall-health self-assessment test result: PASSED > > General SMART Values: > Offline data collection status: (0x84) Offline data collection activity > was suspended by an interrupting > command from host. > Auto Offline Data Collection: Enabled. > Self-test execution status: ( 0) The previous self-test routine > completed > without error or no self-test has ever > been run. > Total time to complete Offline > data collection: (9480) seconds. > Offline data collection > capabilities: (0x7b) SMART execute Offline immediate. > Auto Offline data collection on/off > support. > Suspend Offline collection upon new > command. > Offline surface scan supported. > Self-test supported. > Conveyance Self-test supported. > Selective Self-test supported. > SMART capabilities:(0x0003) Saves SMART data before entering > power-saving mode. > Supports SMART auto save timer. > Error logging capability:(0x01) Error logging supported. > General Purpose Logging supported. > Short self-test routine > recommended polling time: ( 2) minutes. > Extended self-test routine > recommended polling time: ( 112) minutes. > Conveyance self-test routine > recommended polling time: ( 5) minutes. > SCT capabilities:(0x303f) SCT Status supported. > SCT Feature Control supported. > SCT Data Table supported. > > SMART Attributes Data Structure revision number: 16 > Vendor Specific SMART Attributes with Thresholds: > ID# ATTRIBUTE_NAME FLAG VALUE WORST THRESH TYPE UPDATED > WHEN_FAILED RAW_VALUE > 1 Raw_Read_Error_Rate 0x002f 197 197 051Pre-fail Always > - 900 > 3 Spin_Up_Time0x0027 239 236 021Pre-fail Always > - 1050 > 4 Start_Stop_Count0x0032 100 100 000Old_age Always > - 42 > 5 Reallocated_Sector_Ct 0x0033 200 200 140Pre-fail Always > - 0 > 7 Seek_Error_Rate 0x002e 200 200 000Old_age Always > - 0 > 9 Power_On_Hours 0x0032 100 100 000Old_age Always > - 401 > 10 Spin_Retry_Count0x0032 100 253 000Old_age Always > - 0 > 11 Calibration_Retry_Count 0x0032 100 253 000Old_age Always > -
Re: I/O delays
Hi Konstantin, Thank you for the reply. Here is the output of smartctl. We will see if a different drive model helps. Sincerely, Devin [r...@lnx4103 ~]# smartctl -a /dev/sda smartctl version 5.38 [x86_64-redhat-linux-gnu] Copyright (C) 2002-8 Bruce Allen Home page is http://smartmontools.sourceforge.net/ === START OF INFORMATION SECTION === Device Model: WDC WD5002ABYS-02B1B0 Serial Number:WD-WCASYDXX Firmware Version: 02.03B03 User Capacity:500,107,862,016 bytes Device is:Not in smartctl database [for details use: -P showall] ATA Version is: 8 ATA Standard is: Exact ATA specification draft version not indicated Local Time is:Wed Sep 8 23:34:17 2010 EDT SMART support is: Available - device has SMART capability. SMART support is: Enabled === START OF READ SMART DATA SECTION === SMART overall-health self-assessment test result: PASSED General SMART Values: Offline data collection status: (0x84) Offline data collection activity was suspended by an interrupting command from host. Auto Offline Data Collection: Enabled. Self-test execution status: ( 0) The previous self-test routine completed without error or no self-test has ever been run. Total time to complete Offline data collection: (9480) seconds. Offline data collection capabilities:(0x7b) SMART execute Offline immediate. Auto Offline data collection on/off support. Suspend Offline collection upon new command. Offline surface scan supported. Self-test supported. Conveyance Self-test supported. Selective Self-test supported. SMART capabilities:(0x0003) Saves SMART data before entering power-saving mode. Supports SMART auto save timer. Error logging capability:(0x01) Error logging supported. General Purpose Logging supported. Short self-test routine recommended polling time:( 2) minutes. Extended self-test routine recommended polling time:( 112) minutes. Conveyance self-test routine recommended polling time:( 5) minutes. SCT capabilities: (0x303f) SCT Status supported. SCT Feature Control supported. SCT Data Table supported. SMART Attributes Data Structure revision number: 16 Vendor Specific SMART Attributes with Thresholds: ID# ATTRIBUTE_NAME FLAG VALUE WORST THRESH TYPE UPDATED WHEN_FAILED RAW_VALUE 1 Raw_Read_Error_Rate 0x002f 197 197 051Pre-fail Always - 900 3 Spin_Up_Time0x0027 239 236 021Pre-fail Always - 1050 4 Start_Stop_Count0x0032 100 100 000Old_age Always - 42 5 Reallocated_Sector_Ct 0x0033 200 200 140Pre-fail Always - 0 7 Seek_Error_Rate 0x002e 200 200 000Old_age Always - 0 9 Power_On_Hours 0x0032 100 100 000Old_age Always - 401 10 Spin_Retry_Count0x0032 100 253 000Old_age Always - 0 11 Calibration_Retry_Count 0x0032 100 253 000Old_age Always - 0 12 Power_Cycle_Count 0x0032 100 100 000Old_age Always - 40 192 Power-Off_Retract_Count 0x0032 200 200 000Old_age Always - 39 193 Load_Cycle_Count0x0032 200 200 000Old_age Always - 2 194 Temperature_Celsius 0x0022 121 112 000Old_age Always - 26 196 Reallocated_Event_Count 0x0032 200 200 000Old_age Always - 0 197 Current_Pending_Sector 0x0032 200 200 000Old_age Always - 0 198 Offline_Uncorrectable 0x0030 100 253 000Old_age Offline - 0 199 UDMA_CRC_Error_Count0x0032 200 200 000Old_age Always - 0 200 Multi_Zone_Error_Rate 0x0008 100 253 000Old_age Offline - 0 SMART Error Log Version: 1 No Errors Logged SMART Self-test log structure revision number 1 No self-tests have been logged. [To run self-tests, use: smartctl -t] SMART Selective self-test log data structure revision number 1 SPAN MIN_LBA MAX_LBA CURRENT_TEST_STATUS 100 Not_testing 200 Not_testing 300 Not_testing 400 Not_testing 500 Not_testi
Re: I/O delays
Hi Steve, On Sep 8, 2010, at 5:03 PM, Steven Timm wrote: > Devin--what is the output of the command "dmesg" > Are there any kernel traces or bugs in that output? Unfortunately I don't see anything unusual in syslog or the output of dmesg. Thanks, Devin dmesg.log Description: dmesg.log
Re: I/O delays
On Wed, Sep 08, 2010 at 04:42:43PM -0400, Devin Bougie wrote: > Hi, All. We are seeing periodic I/O delays on a new large compute node ... Hi, you do not say what your disks are (model number as reported by smartctl), but I have seen same symptoms as you describe with some WDC 2TB "advanced format" disks. The problem vanished with the next shipment of these 2TB disks, which also happen to have a slightly different model number. So I say, if you see disk delays, try different disks (different maker, different model, different capacity). -- Konstantin Olchanski Data Acquisition Systems: The Bytes Must Flow! Email: olchansk-at-triumf-dot-ca Snail mail: 4004 Wesbrook Mall, TRIUMF, Vancouver, B.C., V6T 2A3, Canada
Re: I/O delays
Devin--what is the output of the command "dmesg" Are there any kernel traces or bugs in that output? Steve On Wed, 8 Sep 2010, Devin Bougie wrote: Hi, All. We are seeing periodic I/O delays on a new large compute node using two Xeon X5670 CPU's (hyper-threaded for a total of 24 processors) with 48GB memory. The system is running the latest kernel on a fully updated x86_64 SL5.5. So far my attempts at changing schedulers and BIOS configurations has not helped. This is a unique system for us, and I have not been able to reproduce this on any other x86_64 SL5.5 systems. Various benchmarks we've run (most extensively using Mathematica) show the system performing very well. So far we've only seen the problem when performing disk I/O. Please let me know if there is any more information I can provide, and any suggestions would be greatly appreciated. Many thanks, Devin -- Here is an example of how we can reproduce the delays by repeatedly writing modest amounts of data. -- [da...@lnx4103 ~]% time dd if=/dev/zero of=/mnt/scratch/test bs=1M count=1K 1024+0 records in 1024+0 records out 1073741824 bytes (1.1 GB) copied, 0.914825 seconds, 1.2 GB/s real0m2.037s user0m0.000s sys 0m1.050s [da...@lnx4103 ~]% time dd if=/dev/zero of=/mnt/scratch/test bs=1M count=1K 1024+0 records in 1024+0 records out 1073741824 bytes (1.1 GB) copied, 0.903858 seconds, 1.2 GB/s real0m1.129s user0m0.001s sys 0m1.125s [da...@lnx4103 ~]% time dd if=/dev/zero of=/mnt/scratch/test bs=1M count=1K 1024+0 records in 1024+0 records out 1073741824 bytes (1.1 GB) copied, 0.923554 seconds, 1.2 GB/s real0m1.141s user0m0.001s sys 0m1.138s [da...@lnx4103 ~]% time dd if=/dev/zero of=/mnt/scratch/test bs=1M count=1K 1024+0 records in 1024+0 records out 1073741824 bytes (1.1 GB) copied, 0.911595 seconds, 1.2 GB/s real1m27.200s user0m0.001s sys 0m1.148s -- Here is a look at "vmstat -S M 5" before, during, and after the above tests. -- procs ---memory-- ---swap-- -io --system-- -cpu-- r b swpd free buff cache si sobibo in cs us sy id wa st 0 0 0 46937 35 118400 481 437 0 0 98 2 0 0 0 0 46937 35 118400 118 1017 85 0 0 100 0 0 0 0 0 46937 35 118400 0 5 1013 93 0 0 100 0 0 0 0 0 46937 35 118400 042 1017 84 0 0 100 0 0 0 1 0 46937 35 118400 1 2 1005 91 0 0 100 0 0 0 0 0 46937 35 118400 017 1016 90 0 0 100 0 0 0 0 0 46937 35 118400 0 5 1012 89 0 0 100 0 0 0 0 0 46937 35 118400 0 5 1010 85 0 0 100 0 0 0 1 0 46919 35 118600 59612 1071 224 0 0 99 1 0 0 0 0 46919 35 118600 070 1019 89 0 0 100 0 0 0 0 0 46919 35 118600 011 1015 98 0 0 100 0 0 0 0 0 46919 35 118600 0 7 1021 112 0 0 100 0 0 1 0 0 47780 3833700 031 1035 760 0 2 98 0 0 0 5 0 46929 38 113500 0 26431 1040 341 0 1 94 5 0 0 6 0 46986 38 108100 2 9840 1032 158 0 0 84 16 0 0 6 0 47054 39 10130065 13113 1052 186 0 0 85 15 0 0 5 0 47137 3994800 2 13006 1036 171 0 0 86 14 0 0 7 0 47208 3988100 0 13135 1035 161 0 0 86 14 0 0 6 0 47258 3983300 0 9736 1031 150 0 0 84 16 0 0 6 0 47300 3979300 0 8656 1025 131 0 0 83 17 0 0 5 0 47349 3975000 0 2 1027 569 0 0 89 11 0 procs ---memory-- ---swap-- -io --system-- -cpu-- r b swpd free buff cache si sobibo in cs us sy id wa st 0 3 0 47437 3966000 0 26781 1042 1372 0 0 88 12 0 0 5 0 47517 3958500 0 13008 1041 177 0 0 85 15 0 0 5 0 47571 3953100 0 13107 1030 148 0 0 83 17 0 0 5 0 47622 3948400 0 9734 1033 150 0 0 87 13 0 0 5 0 47668 3944000 0 6554 1029 134 0 0 87 12 0 0 5 0 47711 3939900 1 9832 1027 146 0 0 83 17 0 0 4 0 47772 3934000 0 13006 1036 149 0 0 85 15 0 0 5 0 47841 3927500 0 13110 1037 164 0 0 86 14 0 0 5 0 47909 3921200 0 6754 1037 310 0 0 87 13 0 0 1 0 46894 39 118600 0 11842 1031 1825 0 1 93 6 0 0 4 0 46892 39 118600 0 27131 1055 104 0 0 95 5 0 0 4 0 46893 39 118600 0 9834 1025 110 0 0 93 6 0 0 5
