diff options
| -rw-r--r-- | issues/gemma/HS-Rat-crashes-gemma.gmi | 67 |
1 files changed, 67 insertions, 0 deletions
diff --git a/issues/gemma/HS-Rat-crashes-gemma.gmi b/issues/gemma/HS-Rat-crashes-gemma.gmi new file mode 100644 index 0000000..57bd451 --- /dev/null +++ b/issues/gemma/HS-Rat-crashes-gemma.gmi @@ -0,0 +1,67 @@ +# Large datasets crash gemma + +Running GEMMA on the NSNIH dataset in Genenetwork sends the server in a tail spin and logs `BUG: soft lockup CPU stuck` messages. This obviously is not great and appears to be a side effect of running openblas aggressively in parallel. + +## Tags + +* assigned: pjotrp, zachs + +## Tasks + +* [ ] tux02: test out-of-band-access +* [ ] tux02: test GEMMA +* [ ] tux02: set overcommit memory on tux02 to 2 (see below) +* [ ] tux02: reboot and reinstate services on tux02 +* [ ] tux02: test GEMMA +* [ ] tux02: try and optimize versions of openblas using -O2 + +And do the same on tux01 (production) + +## Notes + +A 'soft lockup' is defined as a bug that causes the kernel to loop in kernel mode for more than 20 seconds without giving other tasks a chance to run. The watchdog daemon will send an non-maskable interrupt (NMI) to all CPUs in the system who, in turn, print the stack traces of their currently running tasks. + +We see + +``` +[2512382.403215] watchdog: BUG: soft lockup - CPU#118 stuck for 22s! [migration/118:609] +[2512404.477219] Out of memory: Kill process 1723 (gemma) score 87 or sacrifice child +[2512404.569158] Killed process 1723 (gemma) total-vm:44620288kB, anon-rss:25261688kB, file-rss:0kB, shmem-rss:0kB +[2512405.788221] oom_reaper: reaped process 1723 (gemma), now anon-rss:0kB, file-rss:0kB, shmem-rss:0kB +``` + +It is clear parallel GEMMA is running out of RAM. +You can make softlocks messages relax by setting `/proc/sys/kernel/watchdog_thresh` higher. + +Also overcommit is set to 0 on Tux01. We may want to change that to + +``` +vm.overcommit_memory=2 +vm.overcommit_ratio=90 +``` + +That will make out-of-RAM problems less impactful. I have not set that before because it requires rebooting the production server. + +From Zach I got the K and GWA commands: + +``` +/usr/local/guix-profiles/gn-latest-20220122/bin/gemma-wrapper --json --loco 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,X -- -debug -g /home/zas1024/gn2-zach/genotype_files/genotype/bimbam/HSNIH-Palmer_true_geno.txt -p /home/zas1024/gn2-zach/tmp/gn2/gn2/PHENO_2+FcfQiTVSC7FmmbsatUPg.txt -a /home/zas1024/gn2-zach/genotype_files/genotype/bimbam/HSNIH-Palmer_true_snps.txt -gk > /home/zas1024/gn2-zach/tmp/gn2/gn2/HSNIH-Palmer_K_TPTFHJ.json + +/usr/local/guix-profiles/gn-latest-20220122/bin/gemma-wrapper --json --loco --input /home/zas1024/gn2-zach/tmp/gn2/gn2/HSNIH-Palmer_K_TPTFHJ.json -- -debug -g /home/zas1024/gn2-zach/genotype_files/genotype/bimbam/HSNIH-Palmer_true_geno.txt -p /home/zas1024/gn2-zach/tmp/gn2/gn2/PHENO_2+FcfQiTVSC7FmmbsatUPg.txt -a /home/zas1024/gn2-zach/genotype_files/genotype/bimbam/HSNIH-Palmer_true_snps.txt -lmm 9 -maf 0.05 > /home/zas1024/gn2-zach/tmp/gn2/gn2/HSNIH-Palmer_GWA_MWKKYW.json +``` + +The geno file is massive: + +``` +3.7G Mar 12 11:56 HSNIH-Palmer_true_geno.txt + 24K Mar 12 11:56 PHENO_2+FcfQiTVSC7FmmbsatUPg.txt +3.4M Mar 12 11:56 HSNIH-Palmer_true_snps.txt +``` + +Probably best to test on a different machine! Let's move to tux02. Running luna (a half year old version of GN2) gives `**** FAILED: number of columns in the kinship file does not match the number of individuals for row = 79`. So, that does not help! I think this is a known issue that got fixed later. Next up, try and run gemma by hand for chromosome 1 after installing gemma tools with a recent GNU Guix: + +``` + +``` + +Now the goal is to try and crash the server before setting overcommit. |