ecp-veloc / er Goto Github PK
View Code? Open in Web Editor NEWHigh-level distributed erasure coding library combining functionality of shuffile and redset
License: MIT License
High-level distributed erasure coding library combining functionality of shuffile and redset
License: MIT License
See SCR v3.0rc2 project: https://github.com/LLNL/scr/projects/3
See SCR v3.0 project: https://github.com/LLNL/scr/projects/4
See SCR v4.0 project: https://github.com/orgs/LLNL/projects/26
The shuffile file is not correctly restored after rebuilding lost files. It should contain a list of files owned by the rank:
RANK
1
FILE
/dev/shm/user123/scr.1628391/scr.dataset.13/.scr/reddesc.er.1.redset
/dev/shm/user123/scr.1628391/scr.dataset.13/.scr/reddesc.er.1.xor.grp_1_of_1.mem_2_of_2.redset
/dev/shm/user123/scr.1628391/scr.dataset.13/rank_1.ckpt
FILES = 3
But after a rebuild, it only consists of:
RANK = 0
The shuffile file is not part of the set of files protected by the redundancy encoding, so it needs to be restored independently.
It could be reconstructed by reading the list of files from the redset after it has succeeded and before deleting it here:
Lines 737 to 741 in 9e88b5b
The simplest approach would be to just recreate the shuffile file on every rank (even those where it already exists) by calling shuffile_create
again:
Lines 683 to 687 in 9e88b5b
To do that, we need to be able to list the files in the redset. The redset_filelist_get
function only returns redundancy encoding files. We will need to add a new function to redset to return the full file list: user files + redundancy files.
I've packaged er for the gentoo overlay guru but our CI failed at testing it
https://bugs.gentoo.org/784644
Build log: https://784644.bugs.gentoo.org/attachment.cgi?id=701382
Test log: https://784644.bugs.gentoo.org/attachment.cgi?id=701385
1/2 Testing: er_test
1/2 Test: er_test
Command: "/usr/bin/mpirun" "-np" "3" "./er_test" "256"
Directory: /var/tmp/portage/sys-cluster/er-0.0.4/work/er-0.0.4_build/test
"er_test" start time: Apr 21 02:51 CEST
Output:
----------------------------------------------------------
REDSET 0.0.5 ABORT: rank 0 on localhost: Number of partner replicas (1) must be in range [1,1) with 1 ranks in set @ /var/tmp/portage/sys-cluster/redset-0.0.5/work/redset-0.0.5/src/redset_partner.c:116
REDSET 0.0.5 ABORT: rank 1 on localhost: Number of partner replicas (1) must be in range [1,1) with 1 ranks in set @ /var/tmp/portage/sys-cluster/redset-0.0.5/work/redset-0.0.5/src/redset_partner.c:116
REDSET 0.0.5 ABORT: rank 2 on localhost: Number of partner replicas (1) must be in range [1,1) with 1 ranks in set @ /var/tmp/portage/sys-cluster/redset-0.0.5/work/redset-0.0.5/src/redset_partner.c:116
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD
with errorcode -1.
Like redset, the test case will need to be changed to avoid using hostname to define failure groups, if possible.
Main code is causing the issue:
/tmp/veloc/er/src/er.c:322:5: error: ‘REDSET_KEY_CONFIG_DEBUG’ undeclared (first use in this function); did you mean ‘ER_KEY_CONFIG_DEBUG’?
322 | REDSET_KEY_CONFIG_DEBUG, er_debug);
| ^~~~~~~~~~~~~~~~~~~~~~~
See SCR v3.0 project: https://github.com/llnl/scr/projects/2
The ER README.md points users to er.h for usage info, but documentation in that file is currently lacking. We should say which ER functions are collective. For collective functions, we should say which group of processes must call the function, and we should say whether the return code is the same on all procs or not. More description should be provided for input parameters, including adding text showing example calls.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.