I have a 9w Odyssea UV sterilizer on my salt tank, and the first bulb lasted around a year. When it failed, the glass was discolored and one filament was broken. I put a replacement bulb in this past Nov and it did the same thing the other day. Any ideas why the bulbs are not lasting? They are not getting wet or anything. The unit is on 24X7 so they are not subject to a lot of starting. The 9w bulb matches the ballast. I just ordered a replacement bulb. 

This wattage is close to the wattage of the miniature T5s 4-13W.
I don't know what was their lifespan in the past, however today, miniature T5s have an average life of a signle year (8,000) on 24/7 use.
If your lamp is a PL style CGL (Compact Germicidal Lamp), so perhaps the lamp is a cheap one of low quality. Buy high quality disinfection lamp, not a cheap chinese one.
The glass shouldn't be discolor, unlike plastic.

@dor123: On UV lamps the glass does discolor, it is in fact the main cause for their relatively short rated life - not more then 5000 hours, then they have to be replaced (even if they "seems to work"), as the UV output can not be guaranteed (you can not judge it by eyes and usually proper UV level meter is not available) and as mainly in sterilization applications the low UV output mean health hazard.

@static: Check the ballast current (if the ballast is not dying), it should be in the range printed on it. Interturn's shorts inside the choke increase the current and this may overload the tube.

I would try another tube in it, branded this time and see how long that lasts. If it dies again quickly change the ballast!

New Howard ballast is on the way. I did not want to risk killing another bulb.

@dor123: On UV lamps the glass does discolor, it is in fact the main cause for their relatively short rated life - not more then 5000 hours, then they have to be replaced (even if they "seems to work"), as the UV output can not be guaranteed (you can not judge it by eyes and usually proper UV level meter is not available) and as mainly in sterilization applications the low UV output mean health hazard.

@Medved: How glass can discolor? The windows in the buildings and the cars are made from glass and they don't discolors, despite they are exposed to sunlight (Discoloring means losing transparency and/or color changing from transparent or white to yellow or brown from the UV and this can only accurs in plastics).
Quartz is also a type of glass so there is no reason for it to discolor, especially if it don't absorbs the UVC radiation.

@dor123: The UV present at earth surface has rather low energy, as the "harder" UV wavelengths are absorbed by the atmosphere.
Glass is indeed quite robust, but if the energy exceed some limit, so the glass start to absorb it, it usually lead to structure changes (some molecules are broken and eventually reform to other; mainly those added to SiO2 in the glass mix, like sodium compounds) resulting in increased absorbtion.
This starts affect the shortest wavelength first, slowly progressing to longer one.
Actually the visible discoloration i've seen reported was caused by Cobalt gamma emitter.
But by the discoloration i mean altering the transparency even outside the visible range.
And this happen in the tube glass of most sterilization UV lamps: The glass still look transparent in the visible range, but it is not as much transparent anymore in the short wave UV required for the sterilization.

Quartz is way more robust, as it contain only SiO2, so if these break, their atoms may form only again the SiO2, so again the quartz, the only problem is then the crystallization (newly formed molecules tend to assemble in a crystal structure)...
But the consequence of only one component is very high and narrow temperature area, where it is possible to form it, making all product from it rather expensive, so the use of quartz is avoided whenever possible...

I have never had one of the bulbs last long enough to discolor from the UV, they look like something exploded near the filaments. I posted pictures of this in my gallery.

In my case, I think the ballast killed the bulb long before it reached the end of its useful life. The green hair algae started to appear and then I checked the bulb and it was dead (one filament blown)  Hopefully when I get the new ballast the bulb will last long enough for me to see what it looks like when the uv output drops off.

@Static: You would see Algae growing at the same time with the bulb "lighting"...
In these bulbs it would be maximum very light brownish tint, or not visible at all...