Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://www.eso.org/~qc/dfos/known_issues1.html
Дата изменения: Tue Oct 22 12:45:35 2013 Дата индексирования: Thu Feb 27 23:49:51 2014 Кодировка: Поисковые слова: http astrokuban.info astrokuban |
Common DFOS tools:
|
dfos = Data Flow Operations System, the common tool set for DFO |
![]() |
scp from operational machines to qcweb very slow.
Symptoms: scp from operational machines (as used frequently within the dfos tools) apparently gets stuck for seconds/minutes, may even time out. Connections with the web browser to the content on qcweb also time out. Symptoms look like network/connection issue, but they aren't.
Background: When this happened in August/September 2013, qcweb had undergone a change in nfs architecture, causing very inefficient disk I/O. With the many tarballs delivered to qcweb and being unpacked there (in the course of moveProducts: logs, plots), there was in addition processing load on the machine. This caused loads as high as 10 or 20. Even with a hardware upgrade from one to four cores, and additional memory, this load was very high.
Solution: After a simple reboot of the system the hanging processes, and the disk I/O issue did not show up again. See here the monthly/yearly load on qcweb, demonstrating how the issue build up over the two months of August and September 2013, and disappeared thereafter.
![]() |
![]() |