Multiple [RESOLVED]: inotify kernel panic

[Matt]

Following a kernel upgrade on January 23 from 2.6.25 to 2.6.32 to address Ruby 1.9 wrappers, servers have exhibited random lock-ups from what appears to be a bug within the inotify subsystem. Echelon experienced the first outage on January 27, with an outage recorded by Aleph on February 1st.

Kernel updates will continue to be monitored for relevant patches and another maintenance window will be set in the near future to upgrade kernels once again.

* More on this as it develops.

---

2010/02/02: Should either server require a reboot over the next couple weeks, each server has the latest kernel (2.6.32.7) loaded. Aleph will revert to the SLAB allocator, while Echelon continues to use SLUB just in case problems are linked to the recent server optimization.

[Matt]

Marking this as resolved. It has been a month since several services were upgraded addressing potential causes for erratic performance. Uptime since mid-March has been impeccable with no abrupt spikes in server loads, which were symptomatic of the problems mentioned.