I was returning from an 11 hour tour of ten airports on the east side of the Sierra mountains. I had refueled at Apple Valley (APV) and had burned three hours of gas as I was approaching Salinas VOR at 6000 on an IFR clearance to my home airport of Watsonville. Oakland and the Bay Area were reporting 1200 overcast and drizzle, but Watsonville and Monterey reported 1300 overcast and 10 miles visibility. On the moonless night, clouds occassionally appeared at my altitude, but tops of the overcast were reported a believable 2500. Although I could not see Hollister or Paso Robles, there were clear areas behind me at Harris Ranch or eastward (about an hour away), but I could see the glow through the clouds of each of the nearby cities. As I approached SNS VOR, my AI showed I had a nose up pitch of about 10 degrees. Since my power, altimeter, and airspeed all indicated cruise flight (2400 RPM, level altitude, 100 knots), I immediately checked the wings for ice. There was none, so I tried to reset the miniature airplane. As I was fooling with it, I noticed the vaccuum gauge, which I had moved to a place between the AI and DG when I rearranged the panel from its 1966 scatterplot. The gauge showed no pressure. I then checked the DG, and noticed it very slowly start a counterclockwise spin. I called ATC, declined to declare an emergency, and told them my intentions were to do a hold at SNS for a few turns and then continue for the WVI localizer RWY 2. They knew I was solo and had all my information from my flight plan. As the gyros spun down and I flew straight and level for the VOR, the AI slowly showed a steepening climb to the left while the DG started to spin faster showing a steepening turn to the right. In a left turn the DG spin would slow, and during a right turn it spun faster. I always keep a coverup sucked to the left window, so I put it over the AI and took another from my flight bag and covered the DG. I flew two turns in the hold, practicing timed turns to my magnetic vertical card compass. The electric turn and bank seemed quite accurate, and I got eight minutes of valuable practice. I was then cleared for the LOC 2 approach to WVI and headed out towards the IAF, NALLS intersection. As I thought about priorities, I decided that my most important goal was to not exceed a standard rate turn, and not turn more than 30 degrees off course. Tracking the localizer was secondary, since attempts to recapture it would be more dangerous than simply flying in a wings-level descent. I knew that by pulling power off at NALLS and remaining upright, I would descend through the clouds before ever getting to the shore. This meant collision risk with high terrain or churches was negligible. So I crossed NALLS, pulled power off, and set my trim. ATC reminded me it was very important to cancel IFR after landing, and gave me a frequency change. I gave seven clicks for the all important runway lights, and continued the descent. As I continued the approach, I focused on half-standard rate turns. I convinced myself to avoid looking outside until I got to MDA (680). I think I ended up in a little right side slip, which, combined with the wind, got me to a half-deflection by MDA. I looked outside, found the runway, landed safely, and cancelled IFR. Sure enough, the firewall near the outlet from the vaccum pump was covered with black, sooty powder. In retrospect, there were a lot of factors in deciding to finish the approach at WVI instead of flying off to VFR conditions. The most important were fuel, recent experience and training, familiarity with the approach, weather conditions, and pilot condition. I had burned a little over 3 hours of fuel. Night IFR I have found that fuel is extremely important. I was not excited about flying another hour away on a very dark night over high terrain to an unfamiliar airport after 11 hours of flying. On the other hand, flying a no-gyro approach I had practiced only two weeks earlier to my home airport with smooth air, a fairly high ceiling, and no obstacles or populated areas on the approach seemed like a better idea. I was solo, so any mistakes over the water would only hurt me. I didn't declare an emergency for many reasons: because I was solo, because I didn't need any priority handling (I already had all the weather, additional vectors or diversion to a heavily populated Oakland landing were unwise), and also because I didn't want any distracting help from ATC. This last reason was probably the most important factor. ATC may ask me to switch frequencies, change squawk, ask me if I'm still VFR, suggest that Modesto is clear and a good alternate, give me extra traffic alerts, etc. As a CFII, I believe in an emergency I will do very well if I continue to fly as normally as possible, accept the emergency conditions, reduce the distractions and reduce workload. This means to fly the approach you're mentally prepared for, accept the vaccuum failure instead of diagnosing it or trying to fix it, cover up the AI and DG and keep ATC quiet, and focus on the important task of keeping the wings level instead of perfectly tracking the localizer. Other experienced pilots I have spoken to have said they would definitely divert to a VFR airport an hour away. I think there are factors that would make me agree: more gas, daylight, lower ceiling, worse visibility, terrain near the approach, lack of recent no-gyro experience, or guaranteed VFR at the alternate. I certainly would have diverted if I had two hours of gas, weather 500/3, was flying the Salinas ILS into the valley, was marginally current, and knew the diversion airport would remain very clear. I also would have rather done an NDB approach instead of a localizer. The ADF needle gives instantaneous bearing information while the localizer does not. A pilot can simply fly a homing course on the NDB station, and at reasonable distances away, it will give the same rate of turn information as a DG. In other words, from the FAF to the MDA, a pilot on an NDB approach can probably fly the approach with no other instrumentation at all. A pilot on a localizer is hard pressed to fly correctly without at least a turn and bank indicator. From this experience I think by far the best way to survive a vaccuum failure is to recognise it and cover up the AI. My AI OFF flag did not come down during this experience, so I highly recommend ensuring the vaccuum guage is working and installed between the two instruments. I also highly recommend keeping at least one coverup handy at all times. The tumbling AI is extremely distracting otherwise. The coverup was very useful to me on another IFR flight when my ASI failed. I think once the vaccuum failure is recognized, and the AI is covered, most IFR pilots could have landed safely like I did. If I had to shoot an ILS to minimums in turbulence using a turn coordinator, I would not be so confident.